CN112651539A

Movatterモバイル変換

Info

Publication number: CN112651539A
Application number: CN201910967985.8A
Authority: CN
Inventors: 夏迎炬; 马军; 杨铭; 刘汝杰
Original assignee: Fujitsu Ltd
Current assignee: Fujitsu Ltd
Priority date: 2019-10-12
Filing date: 2019-10-12
Publication date: 2021-04-13
Also published as: JP2021064365A

Abstract

Translated fromChinese

公开了一种信息处理装置和信息处理方法。其中，信息处理装置包括：变化点检测单元，被配置成使用第一数据划分单元从对象的历史特性数据序列中检测多个变化点，其中，多个变化点中第一变化点的特性相对于第一数据划分单元中的数据的统计特性的第一变化程度大于多个变化点中第二变化点的特性相对于第一数据划分单元中的数据的统计特性的第二变化程度；历史特性数据序列分割单元，被配置成基于所检测到的第一变化点和第二变化点，将历史特性数据序列分割为多个数据分段；以及训练单元，被配置成基于多个数据分段来训练对象的预测器，其中，预测器用于预测对象的随时间变化的特性数据。

An information processing apparatus and an information processing method are disclosed. Wherein, the information processing apparatus includes: a change point detection unit configured to use a first data dividing unit to detect a plurality of change points from the historical characteristic data sequence of the object, wherein the characteristic of the first change point among the plurality of change points is relative to The first degree of change of the statistical characteristics of the data in the first data division unit is greater than the second degree of change of the characteristics of the second change point in the plurality of change points relative to the second degree of change of the statistical characteristics of the data in the first data division unit; historical characteristic data a sequence segmentation unit configured to segment the historical characteristic data sequence into a plurality of data segments based on the detected first change point and the second change point; and a training unit configured to train based on the plurality of data segments A predictor of an object, wherein the predictor is used to predict the time-varying characteristic data of the object.

Description

Information processing apparatus, information processing method, and computer program

Technical Field

The present disclosure relates to the field of information processing, and more particularly, to an information processing apparatus and an information processing method capable of improving accuracy of a temporal change of a prediction target.

Background

Predicting time-varying data of a subject is a research focus today.

Inventory forecasting is one example of an application that predicts characteristic data of an object over time. The stock serves as a buffer and becomes a reservoir of the enterprise. It is the stock of raw materials, semi-finished products and finished products made in advance in order to complete production tasks in the enterprise operation process. Inventory exists as a means of enterprise production continuity and stability and to cope with uncertain factors such as sudden increases in consumer demand, sudden increases in the price of raw materials produced, etc. Excessive inventory masks many of the issues in enterprise development. On one hand, the inventory cost accounts for a large part of the total enterprise cost, and on the other hand, the inventory amount directly influences the fund flow of the enterprise, the overall efficiency and the market core competitiveness.

Inventory management also has the disadvantage of being difficult to deal with emergencies, and once a supply chain is damaged or an enterprise cannot adjust production according to customer requirements in a short time, the stability of production and operation of the enterprise is affected, and the operation risk is increased. To ensure frequent small deliveries under contractual agreement, the supplier may require additional pricing, and the business therefore loses the opportunity to earn lower prices from other suppliers. In any link, errors occur, the whole business chain is possibly disordered, and under the condition that the commodity demand is suddenly increased, the enterprise is short of goods and long in supply time, so that great opportunity is lost, and the enterprise is seriously lost, therefore, the significance of understanding the inventory management lies in formulating a proper inventory management strategy according to the characteristics of the enterprise.

With the support of artificial intelligence, the inventory management can optimize each link of the inventory management by using a machine learning technology on the basis of big data. The method comprises the steps of firstly forecasting the demand of the inventory and optimizing inventory management on the basis of forecasting.

In the prior art, traditional prediction methods such as statistical prediction methods and experience-based prediction methods are used to assist in decision making. Conventional prediction methods typically do not achieve accurate predictions.

Disclosure of Invention

The following presents a simplified summary of the disclosure in order to provide a basic understanding of some aspects of the disclosure. However, it should be understood that this summary is not an exhaustive overview of the disclosure. It is not intended to identify key or critical elements of the disclosure or to delineate the scope of the disclosure. Its sole purpose is to present some concepts of the disclosure in a simplified form as a prelude to the more detailed description that is presented later.

In view of the above problems, it is an object of the present disclosure to provide an information processing apparatus and an information processing method capable of solving one or more disadvantages in the related art.

According to an aspect of the present disclosure, there is provided an information processing apparatus including: a change point detecting unit configured to detect a plurality of change points from a history characteristic data sequence of the object using the first data dividing unit, wherein a first degree of change of a characteristic of a first change point among the plurality of change points with respect to a statistical characteristic of data in the first data dividing unit is greater than a second degree of change of a characteristic of a second change point among the plurality of change points with respect to the statistical characteristic of data in the first data dividing unit; a history characteristic data sequence dividing unit configured to divide the history characteristic data sequence into a plurality of data segments based on the detected first change point and second change point; and a training unit configured to train a predictor of the object based on the plurality of data segments, wherein the predictor is used for predicting the time-varying characteristic data of the object.

According to another aspect of the present disclosure, there is provided an information processing method including: a change point detection step of detecting a plurality of change points from a history characteristic data sequence of the object using a first data dividing unit, wherein a first degree of change in a characteristic of a first change point among the plurality of change points with respect to a statistical characteristic of data in the first data dividing unit is larger than a second degree of change in a characteristic of a second change point among the plurality of change points with respect to the statistical characteristic of data in the first data dividing unit; a historical characteristic data sequence dividing step of dividing the historical characteristic data sequence into a plurality of data segments based on the detected first change point and second change point; and training a predictor of the object based on the plurality of data segments, wherein the predictor is used for predicting the characteristic data of the object changing along with time.

According to another aspect of the present disclosure, there is provided an information processing apparatus including a trained predictor obtained according to the above-described information processing device, wherein the trained predictor is configured to predict time-varying characteristic data of an object to be processed.

According to other aspects of the present disclosure, there is also provided computer program code and computer program products for implementing the methods according to the present disclosure.

Additional aspects of the disclosed embodiments are set forth in the description section that follows, wherein the detailed description is presented to fully disclose the preferred embodiments of the disclosed embodiments without imposing limitations thereon.

Drawings

The disclosure may be better understood by reference to the following detailed description taken in conjunction with the accompanying drawings, in which like or similar reference numerals are used throughout the figures to designate like or similar components. The accompanying drawings, which are incorporated in and form a part of the specification, further illustrate preferred embodiments of the present disclosure and explain the principles and advantages of the present disclosure, are incorporated in and form a part of the specification. Wherein:

fig. 1 is a block diagram showing a functional configuration example of an information processing apparatus according to an embodiment of the present disclosure;

2A-2C are diagrams illustrating examples of first change points and/or second change points according to embodiments of the present disclosure;

fig. 3 is a diagram showing an example of dividing a history characteristic data sequence based on a first change point and a second change point according to an embodiment of the present disclosure;

fig. 4 is a flowchart showing an example of a flow of an information processing method according to an embodiment of the present disclosure; and

fig. 5 is a block diagram showing an example structure of a personal computer employable in the embodiments of the present disclosure.

Detailed Description

Exemplary embodiments of the present disclosure will be described hereinafter with reference to the accompanying drawings. In the interest of clarity and conciseness, not all features of an actual implementation are described in the specification. It will of course be appreciated that in the development of any such actual embodiment, numerous implementation-specific decisions must be made to achieve the developers' specific goals, such as compliance with system-related and business-related constraints, which will vary from one implementation to another. Moreover, it will be appreciated that such a development effort might be complex and time-consuming, but would nevertheless be a routine undertaking for those of ordinary skill in the art having the benefit of this disclosure.

Here, it should be further noted that, in order to avoid obscuring the present disclosure with unnecessary details, only the device structures and/or processing steps closely related to the scheme according to the present disclosure are shown in the drawings, and other details not so relevant to the present disclosure are omitted.

Embodiments according to the present disclosure are described in detail below with reference to the accompanying drawings.

First, a functional block diagram of aninformation processing apparatus 100 according to an embodiment of the present disclosure will be described with reference to fig. 1. Fig. 1 is a block diagram showing a functional configuration example of aninformation processing apparatus 100 according to an embodiment of the present disclosure. As shown in fig. 1, theinformation processing apparatus 100 according to the embodiment of the present disclosure includes a change point detection unit 102, a historical characteristic data sequence division unit 104, and a training unit 106.

The change point detecting unit 102 may be configured to detect a plurality of change points from the historical characteristic data sequence of the object using the first data dividing unit, wherein a first degree of change of a characteristic of a first change point of the plurality of change points with respect to a statistical characteristic of data in the first data dividing unit is greater than a second degree of change of a characteristic of a second change point of the plurality of change points with respect to the statistical characteristic of data in the first data dividing unit.

As an example, the object may be various parts involved in an automobile part supply service, the historical characteristic data of the object may be stock quantities of the various parts, and the historical characteristic data may be stock quantities of the various parts at different times. In addition, the object may be a commodity or the like provided by an electronic product supplier, the historical characteristic data of the object may be an inventory quantity of the commodity or the like, and the historical characteristic data may be an inventory quantity of the commodity or the like at different times. Hereinafter, for the sake of brevity, description will be made taking an object as an automobile part as an example.

As an example, the historical characteristic data sequence of the object is a time sequence of the historical characteristic data of the object. For example, the historical characteristic data sequence of the object may be represented as D ═ D₁,d₂,…,d_nIn which d is₁,d₂,…,d_nRespectively represent the 1 st characteristic data, the 2 nd characteristic data, …, the nth characteristic data at various time points in the sequence, wherein the unit of time can be hour, day, month, year, etc.

As an example, the characteristic of the first change point may be a data amplitude of the first change point and the characteristic of the second change point may be a data amplitude of the second change point. The statistical characteristic of the data in the first data dividing unit may be a statistical characteristic value of the history characteristic data included in the first data dividing unit, for example, the statistical characteristic value may be a mean, a variance, a median, a quartile, or the like of the history characteristic data included in the first data dividing unit. Other examples of the characteristics of the first change point, the characteristics of the second change point, and the statistical characteristics of the data in the first data dividing unit may also be conceivable by those skilled in the art, and will not be described here in a repeated manner.

As an example, the first change point includes a plurality of first change points and the second change point includes a plurality of second change points. As an example, the first change point and the second change point may be noise data present in the history characteristic data sequence. The first change point and the second change point may be other data that reflects a change in the historical characteristic data sequence, and may be abnormal data due to factors such as stoppage, promotion, and recall of an object (for example, various parts related to the above-described automobile part supply service, a product provided by an electronic product supplier, and the like). Wherein the first change point is characteristic data whose characteristic has a large degree of change with respect to the statistical characteristic of the data in the first data dividing unit, and the second change point is characteristic data whose characteristic has a small degree of change with respect to the statistical characteristic of the data in the first data dividing unit. As an example, in the case where both the first change point and the second change point are noise data, the first change point may be larger noise data, and the second change point may be smaller noise data; in the case where both the first change point and the second change point are abnormal data, the first change point may be abrupt change data in the historical characteristic data sequence, and the second change point may be gradual change data in the historical characteristic data sequence. These noisy and anomalous data can affect the accuracy of the prediction of the object over time.

Other examples of the first and second change points will also occur to those skilled in the art and will not be discussed here in a repeated fashion.

The change point detecting unit 102 according to the disclosed embodiment can not only detect a change point from the historical characteristic data sequence, but also distinguish the change point into a first change point having a larger degree of change and a second change point having a smaller degree of change, which can reflect the degree of change of data in the historical characteristic data sequence more accurately. In the case where the object is various parts related to an automobile part supply service, the change point detection unit 102 according to the disclosed embodiment can reflect the degree of change of data in the historical inventory data sequence of the automobile parts more accurately by dividing the change point into a first change point with a large degree of change and a second change point with a small degree of change, so that the inventory prediction is more accurate.

Preferably, the change point detecting unit 102 may be configured to slide the first data dividing unit over the history characteristic data sequence and detect the history characteristic data at a predetermined position outside the first data dividing unit as the first change point or the second change point in a case where the history characteristic data satisfies a predetermined condition.

As an example, the first data dividing unit may be a first sliding window, and furthermore, other implementation examples of the first data dividing unit may be also conceivable by those skilled in the art, and will not be described here in a repeated manner. Hereinafter, for brevity, the first data division unit is described as an example of a first sliding window.

Assuming that the width of the first sliding window is W, the first sliding window is made to be from the 1 st characteristic data D in the historical characteristic data sequence D₁The sliding is started. At the beginning of the first sliding window is characteristic data d_i(i ═ 1,2, …, n), the historical characteristic data included in the first sliding window may be represented as D_w＝{d_i,d_i+1,…,d_i+w-1}。

The first change point or the second change point is detected by comparing a characteristic of the historical characteristic data outside the first sliding window with a statistical characteristic of the historical characteristic data within the first sliding window. As an example, the history characteristic data at the predetermined position outside the first data dividing unit may be characteristic data in the history characteristic data sequence at a predetermined distance from the end point of the first sliding window, for example, for the sake of brevity, the first history characteristic data after the end point of the first sliding window may be considered as 1, the second history characteristic data after the end point of the first sliding window may be considered as 2, and the third history characteristic data after the end point of the first sliding window may be considered as 3, …, where the predetermined distance may be set by a person skilled in the art based on experience. As an example, the history characteristic data at a predetermined position outside the first data dividing unit is the first history characteristic data d after the end point of the first sliding window in the history characteristic data series_i+wIn case of d, if_i+wRelative to D_w＝{d_i,d_i+1,…,d_i+w-1System of historical characteristic data inIf the meter characteristic satisfies the predetermined condition, the historical characteristic data d is obtained_i+wDetected as a first point of change or a second point of change.

As an example, each characteristic data in the historical characteristic data sequence may be sequentially detected to determine whether it is a first change point or a second change point; alternatively, the characteristic data in the historical characteristic data sequence may be detected in predetermined steps, and it is determined whether the detected characteristic data is the first change point or the second change point; alternatively, the characteristic data in only a partial region in the historical characteristic data sequence may be detected.

The change point detecting unit 102 according to the disclosed embodiment can more accurately detect the history characteristic data to be detected as the first change point or the second change point based on the statistical characteristics of the history characteristic data included in the first data dividing unit, which is located before the history characteristic data to be detected.

Preferably, the change point detecting unit 102 may be configured to define the predetermined condition using a quartile of the historical characteristic data in the first data dividing unit, wherein the first degree of change and the second degree of change are degrees of change in the magnitudes of the first change point and the second change point, respectively, with respect to the magnitude determined based on the quartile.

As an example, a lower quartile (first quartile) Q based on historical characteristic data in a first sliding window_lAnd upper quartile (third quartile) Q_hTo define the predetermined condition.

For the sake of simplicity, the historical characteristic data which is still at the preset position outside the first data dividing unit is the first historical characteristic data d after the end point of the first sliding window in the historical characteristic data sequence_i+wThe description is made for the sake of example.

In the presence of a catalyst satisfying d_i+w>Q_h+β*(Q_h-Q_l) Or d_i+w<Q_l-β*(Q_h-Q_l) Under predetermined conditions, the history characteristic data d_i+wDetected as a first point of change.

At the point of satisfying Q_h+β*(Q_h-Q_l)>d_i+w>Q_h+α*(Q_h-Q_l) Or Q_l-β*(Q_h-Q_l)<d_i+w<Q_l-α*(Q_h-Q_l) Under predetermined conditions, the history characteristic data d_i+wDetected as a second point of change.

Where α and β are parameters for controlling the first degree of change and the second degree of change, β is larger than α because the first degree of change is larger than the second degree of change as described above. The values of α and β can be set empirically by those skilled in the art, and as an example, α ═ 2 and β ═ 3 can be set.

Since the quartile of the historical characteristic data in the first data dividing unit can reflect not only the numerical distribution of the historical characteristic data but also the positional distribution of the historical characteristic data, the statistical characteristics of the historical characteristic data in the first data dividing unit can be more accurately reflected.

It will be appreciated by those skilled in the art that the predetermined condition may also be defined based on the lower quartile and the upper quartile in a different manner than the above example, or may be defined based on other quartiles of the historical characteristic data in the first sliding window, which will not be reiterated herein.

The history characteristic data at a predetermined position outside the first data dividing unit may be characteristic data in the history characteristic data sequence which is a predetermined distance from the end point of the first sliding window after the end point, and the history characteristic data d_i+wWhether it is the first change point or the second change point is similarly detected, and a description thereof will not be repeated.

As an example, the above-described predetermined condition may also be defined according to a mean, a variance, a median, and the like of the history characteristic data in the first data dividing unit, which will not be described in detail here.

Fig. 2A-2C are graphs showing examples of first and/or second change points, where the data in fig. 2A-2C is derived from certain automotive part inventory data, with the horizontal axis of each graph representing time (e.g., "201606" for 2016 6 months, "201511" for 201511 months) and the vertical axis representing the quantity of inventory of a certain automotive part, according to an embodiment of the disclosure.

In particular, fig. 2A shows a second point of change in the inventory data of the automobile parts "dashboard clips" at different times. In fig. 2A, there is only one second change point without the first change point, and specifically, the intersection point of the vertical dotted line and the curve in fig. 2A is the second change point. As shown in fig. 2A, before and after the second change point, the trend of the data does not change significantly (both are oscillation-up), but the range of the data changes. Before the second change point, the data oscillates and rises within the range of 0 to 150, and after the second change point, the data oscillates and rises within the range of 50 to 300.

FIG. 2B illustrates a first point of change in the inventory data for "premium synthetic engine oil SN 0W-40" at different times. In fig. 2B, there is only one first change point without a second change point, and specifically, the first change point is an intersection point of a vertical solid line with an abscissa of about "201606" and a curve in fig. 2B. As shown in fig. 2B, before and after the first change point, not only the range of data changes, but also the trend of data changes, and before the first change point, the data gently fluctuates within 50, and after the first change point, the range of data change gradually expands to 500 to 3000, and the data tends to oscillate upward.

Fig. 2C shows a first change point and a second change point in the stock data of the automobile part "right front fender outer panel" at different times. In fig. 2C, there are one first change point and two second change points, and specifically, the first change point is an intersection of a solid vertical line having an abscissa of about "201511" and a curved line in fig. 2C, and the two second change points are intersections of two dashed vertical lines and the curved line, respectively.

Above, for purposes of brevity and clarity, fig. 2A shows that only the second change point exists in the inventory data sequence and fig. 2B shows that only the first change point exists in the inventory data sequence, but in actual practice, both the first change point and the second change point typically exist in the inventory data.

The historical characteristic data sequence dividing unit 104 may be configured to divide the historical characteristic data sequence into a plurality of data segments based on the detected first change point and second change point.

The presence of noisy data and/or anomalous data in the historical characteristic data sequence can interfere with the prediction of the time-varying object. For example, in the case where the object is various parts involved in an automobile parts supply service, noise data and/or abnormal data present in a historical inventory data sequence of automobile parts may interfere with the prediction of inventory needs of the automobile parts. The historical characteristic data sequence segmentation unit 104 according to the embodiment of the present disclosure segments the historical characteristic data sequence based on the detected first change points and second change points, thereby facilitating removal or suppression of noise data and/or abnormal data and the like. For example, in the case where the object is an automobile part, the historical characteristic data series segmentation unit 104 according to the embodiment of the present disclosure helps to remove or suppress noise data and/or abnormal data and the like in the historical inventory data series of the automobile part.

Preferably, the historical characteristic data sequence dividing unit 104 may be configured to search for a division point for dividing the historical characteristic data sequence from the detected first change point and second change point using the second data dividing unit to slide on the historical characteristic data sequence, wherein the division point is not included in any of the plurality of data segments.

As an example, the second data dividing unit may be the same as or different from the first data dividing unit.

As an example, the second data dividing unit may be a second sliding window, and furthermore, other implementation examples of the second data dividing unit may be also conceivable by those skilled in the art, and will not be described here in a repeated manner. Hereinafter, for brevity, the second data division unit is described as an example of a second sliding window.

As an example, the window size of the second sliding window may be the same as or different from the window size of the first sliding window. As an example, the window sizes of the first and second sliding windows may be determined based on characteristics of the part. For example, where the part is a periodic part, the window size is determined according to the size of the part's business period. For example, if the production and distribution cycle of parts is 6 months, "6" is used as the window size of the first sliding window and the window size of the second sliding window in the case of "month" as a time unit (for example, the unit of the horizontal axis in fig. 2A to 2C is "month"). In the case of a part without a significant business cycle, the historical property data sequence may also be simply sliced using a one-year span, resulting in a window size of "12" for the second sliding window.

As an example, the historical characteristic data sequence dividing unit 104 divides the historical characteristic data sequence into a plurality of data segments using the found division point, wherein the division point is not included in any of the plurality of data segments. As an example, any one of the characteristic data other than the above-described division point in the historical characteristic data sequence is included in one of the plurality of data segments.

The historical characteristic data sequence dividing unit 104 according to the embodiment of the present disclosure removes the division point from the historical characteristic data sequence by excluding the division point in any of the plurality of data segments, thereby removing or suppressing noise data and/or abnormal data and the like in the historical characteristic data sequence.

Preferably, the historical characteristic data sequence dividing unit 104 may be configured to take the end of the historical characteristic data sequence as the initial starting point of the second data dividing unit, and to slide the second data dividing unit in the direction from the end to the starting point of the historical characteristic data sequence. That is, the history characteristic data sequence dividing unit 104 divides the history characteristic data sequence into a plurality of data segments in a direction from the end of the history characteristic data sequence (which corresponds to the temporally latest history characteristic data in the history characteristic data sequence) to the start (which corresponds to the temporally earliest history characteristic data in the history characteristic data sequence). As an example, only a predetermined region from the end of the historical characteristic data sequence in the historical characteristic data sequence may be segmented (i.e., not all sequences from the end to the beginning of the historical characteristic data sequence) and the plurality of data segments thus obtained are data segments including temporally latest historical characteristic data.

As an example, the historical characteristic data sequence dividing unit 104 may be further configured to take the beginning of the historical characteristic data sequence as the initial starting point of the second data dividing unit, and to slide the second data dividing unit in the direction from the beginning to the end of the historical characteristic data sequence. That is, the historical characteristic data sequence dividing unit 104 divides the historical characteristic data sequence into a plurality of data segments in a direction from the beginning to the end of the historical characteristic data sequence.

Preferably, the historical characteristic data sequence dividing unit 104 may be configured to, when the first change point is found in the second data dividing unit or at the end point of the second data dividing unit, take the found first change point as the dividing point, and move the start point of the second data dividing unit to the found first change point to continue sliding on the historical characteristic data sequence.

As described above, since the degree of change in the characteristic of the first change point with respect to the statistical characteristic of the data in the first sliding window is large, that is, the degree of change in the characteristic of the first change point with respect to the statistical characteristic of the data in the vicinity thereof is large, the interference of the first change point with the prediction of the temporal change of the object is large. Therefore, as long as the first change point is found in the second sliding window or at the end point of the second sliding window, the found first change point is taken as the division point, thereby removing the found first change point from the historical characteristic data sequence. Then, the starting point of the second sliding window is moved to the searched first change point to continue sliding on the historical characteristic data sequence, so as to continue searching the next segmentation point. As an example, when there are a plurality of first change points in the second sliding window, the plurality of first change points are all taken as division points, and the start point of the second sliding window may be moved to the temporally first change point of the plurality of first change points, and then the second sliding window is continued to slide on the history characteristic data sequence to continue to find the next division point.

Preferably, the historical characteristic data sequence dividing unit 104 may be configured to take the second change point as the end point as the dividing point when the end point of the second data dividing unit is the second change point, and move the start point of the second data dividing unit to the second change point as the end point to continue sliding on the historical characteristic data sequence.

As described above, since the second degree of change of the characteristic of the second change point with respect to the statistical characteristic of the data in the first sliding window is smaller than the first degree of change of the characteristic of the first change point with respect to the statistical characteristic of the data in the first sliding window, that is, the degree of change of the characteristic of the second change point with respect to the statistical characteristic of the data in the vicinity thereof is small, the interference of the second change point with the prediction of the temporal change of the object may sometimes not be large. Therefore, only when the end point of the second sliding window is the second change point, the second change point is taken as the division point, thereby removing the second change point from the history characteristic data sequence. Then, the start point of the second sliding window is moved to the second change point to continue sliding on the historical characteristic data sequence to continue searching for the next segmentation point.

Preferably, the historical characteristic data sequence dividing unit 104 may be configured to slide the second data dividing unit by a predetermined step size when the first change point is not found in the second data dividing unit or at the end point of the second data dividing unit and the end point of the second data dividing unit is not the second change point. That is, if the first change point is not found in the second sliding window or at the end point of the second sliding window and the end point of the second sliding window is not the second change point, the second sliding window is slid by a predetermined step length to find the division point. Wherein the predetermined step size can be preset empirically by a person skilled in the art.

Fig. 3 is a diagram showing an example of dividing the historical characteristic data sequence based on the first change point and the second change point according to the embodiment of the present disclosure, in which the data in fig. 3 is derived from historical stock data of an automobile part "windshield outer shield bar No. 3", and the horizontal axis of fig. 3 represents time and the vertical axis represents the stock number of the part.

In fig. 3, there are one first change point and six second change points, where the intersection point of the solid vertical line having the abscissa of about "201204" in fig. 3 with the curved line is the first change point, and the intersection points of the six dashed vertical lines with the curved line are the six second change points, respectively. In the process of determining the division points, the second sliding window starts from the end to the beginning of the curve shown in fig. 3, and it is assumed that two division points (in fig. 3, the two division points are identified by hollow five-pointed stars) are found from the first change point and the six second change points in the finding manner, and as can be seen from fig. 3, the two division points include a second change point (referred to as the first division point for short) and a first change point (referred to as the second division point for short). Using the first and second cut points, the curve shown in fig. 3 can be segmented into the following three data segments: a first data segment from the end of the curve to the first cut point, a second data segment from the first cut point to the second cut point, and a third data segment from the second cut point to the beginning of the curve. As described above, the first division point and the second division point are not included in any one of the three data segments described above.

The training unit 106 may be configured to train a predictor of the subject based on the plurality of data segments obtained by the historical characteristic data sequence segmentation unit 104, wherein the predictor is used to predict the time-varying characteristic data of the subject.

Since the plurality of data segments includes less noise data and/or outlier data than the original historical characteristic data sequence, the training unit 106 according to embodiments of the present disclosure may train the predictor with higher quality data, thereby enabling improved performance of the predictor. Where the object is an automobile part, the training unit 106 according to embodiments of the present disclosure may utilize the higher quality data to predict the inventory requirements of the automobile part.

Preferably, the training unit 106 may be configured to train the predictor separately using each of the plurality of data segments, thereby improving the prediction accuracy of the predictor.

As an example, for a certain automotive part, a predictor may be trained for the automotive part using each of a plurality of data segments. Preferably, several data segments, including the most recent in time data segment, may be used to train predictors for the part separately.

Preferably, the training unit 106 may be configured to cluster a plurality of data segments, and train the predictor using the clustered data segments, thereby further improving the prediction accuracy of the predictor.

As an example, for a certain automobile part, when the data volume of the data segment that is latest in time is small, a clustering method may be used to train a predictor for the part by aggregating data segments at other times similar to the data segment that is latest in time as historical data, so as to train the predictor with more data. Alternatively, after the historical characteristic data sequence is divided, if the historical data amount of a single automobile part is reduced, the data segments of a plurality of automobile parts can be clustered together, and the obtained cluster data is used for training the predictor for the part, so that the predictor can be trained by using more data.

As is apparent from the above description, theinformation processing apparatus 100 according to the embodiment of the present disclosure can reflect the degree of change of data in the history characteristic data sequence more accurately by distinguishing the change point in the history characteristic data sequence into the first change point having a larger degree of change and the second change point having a smaller degree of change; segmenting the original historical characteristic data sequence based on the detected first change point and the second change point, thereby obtaining a plurality of data segments with higher quality compared with the original historical characteristic data sequence; and training the predictor by using the obtained plurality of data segments, the performance of the predictor can be improved. In the case where the object is various parts involved in an automobile part supply service, theinformation processing apparatus 100 according to the embodiment of the present disclosure can more accurately reflect the degree of change of data in the historical stock data sequence of automobile parts; segmenting the original inventory data sequence based on the detected first change point and the second change point, thereby obtaining a plurality of data segments with higher quality compared with the original inventory data sequence; and by predicting the inventory requirements of the automotive parts using the plurality of data segments, inventory management can be optimized on a predictive basis.

Correspondingly to the embodiment of the information processing device, the disclosure also provides an embodiment of an information processing method.

Fig. 4 is a flowchart illustrating an example of a flow of aninformation processing method 400 according to an embodiment of the present disclosure.

As shown in fig. 4, theinformation processing method 400 according to the embodiment of the present disclosure includes a change point detection step S402, a historical characteristic data sequence segmentation step S404, and a training step S406.

Theinformation processing method 400 according to the embodiment of the present disclosure starts at S401.

In the change point detecting step S402, a plurality of change points are detected from the historical characteristic data sequence of the object using the first data dividing unit, wherein a first degree of change in the characteristic of a first change point among the plurality of change points with respect to the statistical characteristic of the data in the first data dividing unit is greater than a second degree of change in the characteristic of a second change point among the plurality of change points with respect to the statistical characteristic of the data in the first data dividing unit.

As an example, the object may be various parts involved in an automobile part supply service, the historical characteristic data of the object may be stock quantities of the various parts, and the historical characteristic data may be stock quantities of the various parts at different times.

Examples of the historical characteristic data sequence on the object, the characteristic of the first change point, the characteristic of the second change point, and the statistical characteristic of the data in the first data dividing unit may be referred to the description on the change point detecting unit 102 in the apparatus embodiment, and the description is not repeated here.

In the change point detection step S402, not only the change point can be detected from the historical characteristic data sequence, but also the change point can be divided into a first change point with a large degree of change and a second change point with a small degree of change, so that the degree of change of the data in the historical characteristic data sequence can be reflected more accurately. In the case where the object is various parts related to the automobile part supply service, the change point detection step S402 divides the change point into a first change point having a large degree of change and a second change point having a small degree of change, thereby reflecting the degree of change of data in the historical inventory data series of the automobile parts more accurately and making the inventory prediction more accurate.

Preferably, in the change point detecting step S402, the first data dividing unit is slid on the history characteristic data sequence, and in a case where the history characteristic data at a predetermined position outside the first data dividing unit satisfies a predetermined condition, the history characteristic data is detected as the first change point or the second change point.

As an example, the first data dividing unit may be a first sliding window, and furthermore, other implementation examples of the first data dividing unit may be also conceivable by those skilled in the art, and will not be described here in a repeated manner.

For an example of the first sliding window, reference may be made to the description of the change point detecting unit 102 in the apparatus embodiment, and the description will not be repeated here.

In the change point detecting step S402, based on the statistical characteristics of the history characteristic data included in the first data dividing unit, which is located before the history characteristic data to be detected, the history characteristic data to be detected can be detected as the first change point or the second change point more accurately.

Preferably, in the change point detecting step S402, the predetermined condition is defined by a quartile of the history characteristic data in the first data dividing unit, wherein the first degree of change and the second degree of change are degrees of change in the magnitudes of the first change point and the second change point, respectively, with respect to the magnitude determined based on the quartile.

Examples of defining the predetermined condition using the quartile of the history characteristic data in the first data dividing unit may be referred to the description of the change point detecting unit 102 in the apparatus embodiment, and the description is not repeated here.

In the historical characteristic data sequence dividing step S404, the historical characteristic data sequence is divided into a plurality of data segments based on the detected first change point and second change point.

In the historical characteristic data sequence dividing step S404, the historical characteristic data sequence is sliced based on the detected first change points and second change points, thereby facilitating removal or suppression of noise data and/or abnormal data and the like. For example, in the case where the object is an automobile part, it is helpful to remove or suppress noise data and/or abnormal data and the like in the historical stock data sequence of the automobile part.

Preferably, in the historical characteristic data sequence dividing step S404, a dividing point for dividing the historical characteristic data sequence is searched from the detected first change point and second change point using the second data dividing unit to slide on the historical characteristic data sequence, wherein the dividing point is not included in any of the plurality of data segments.

As an example, the second data dividing unit may be a second sliding window, and furthermore, other implementation examples of the second data dividing unit may be also conceivable by those skilled in the art, and will not be described here in a repeated manner.

The description of the second sliding window may refer to the description of the historical characteristic data sequence partitioning unit 104 in the apparatus embodiment, and the description is not repeated here.

In the historical characteristic data sequence dividing step S404, by excluding the dividing point in any of the plurality of data segments, the dividing point is removed from the historical characteristic data sequence, thereby removing or suppressing noise data, abnormal data, or the like in the historical characteristic data sequence.

Preferably, in the historical characteristic data sequence dividing step S404, the end of the historical characteristic data sequence is used as the initial starting point of the second data dividing unit, and the second data dividing unit is slid in the direction from the end to the starting point of the historical characteristic data sequence. That is, in the history characteristic data sequence dividing step S404, the history characteristic data sequence is divided into a plurality of data segments in a direction from the end of the history characteristic data sequence (which corresponds to the temporally latest history characteristic data in the history characteristic data sequence) to the start (which corresponds to the temporally earliest history characteristic data in the history characteristic data sequence). As an example, only a predetermined region from the end of the historical characteristic data sequence in the historical characteristic data sequence may be segmented (i.e., not all sequences from the end to the beginning of the historical characteristic data sequence) and the plurality of data segments thus obtained are data segments including the latest historical characteristic data.

As an example, in the historical characteristic data sequence dividing step S404, the second data dividing unit may be slid in the direction from the beginning to the end of the historical characteristic data sequence with the beginning of the historical characteristic data sequence as the initial starting point of the second data dividing unit. That is, in the historical characteristic data sequence dividing step S404, the historical characteristic data sequence is divided into a plurality of data segments in a direction from the beginning to the end of the historical characteristic data sequence.

Preferably, in the historical characteristic data sequence dividing step S404, when the first change point is found in the second data dividing unit or at the end point of the second data dividing unit, the found first change point is taken as the dividing point, and the start point of the second data dividing unit is moved to the found first change point to continue sliding on the historical characteristic data sequence.

Preferably, in the history characteristic data sequence dividing step S404, when the end point of the second data dividing unit is the second change point, the second change point as the end point is made the dividing point, and the moving of the start point of the second data dividing unit to the second change point as the end point continues the sliding on the history characteristic data sequence.

Preferably, in the historical characteristic data sequence dividing step S404, when the first change point is not found in the second data division unit or at the end point of the second data division unit and the end point of the second data division unit is not the second change point, the second data division unit is slid by a predetermined step. That is, if the first change point is not found in the second sliding window or at the end point of the second sliding window and the end point of the second sliding window is not the second change point, the second sliding window is slid by a predetermined step length to find the division point. Wherein the predetermined step size can be preset empirically by a person skilled in the art.

As for an example of dividing the historical characteristic data sequence based on the first change point and the second change point, reference may be made to the description of the historical characteristic data sequence dividing unit 104 and fig. 3 in the apparatus embodiment, and the description will not be repeated here.

In a training step S406, a predictor of the subject is trained based on the plurality of data segments obtained in the historical characteristic data sequence dividing step S404, wherein the predictor is used for predicting the characteristic data of the subject changing with time.

Since the plurality of data segments includes less noise data and/or outlier data than the original historical characteristic data sequence, the predictor can be trained with higher quality data, thereby enabling improved performance of the predictor. In the case where the object is an automobile part, the inventory requirement of the automobile part can be predicted using the data of higher quality.

Preferably, in the training step S406, the predictor is trained separately using each of the plurality of data segments, so as to improve the prediction accuracy of the predictor.

Preferably, in the training step S406, the plurality of data segments are clustered, and the clustered data segments are used to train the predictor, so as to improve the prediction accuracy of the predictor.

Examples of using data segments to train a predictor can be found in the description of the apparatus embodiment with respect to the training unit 106, and will not be repeated here.

Theinformation processing method 400 according to the embodiment of the present disclosure ends at S407.

As is apparent from the above description, theinformation processing method 400 according to the embodiment of the present disclosure can reflect the degree of change of data in the historical characteristic data sequence more accurately by distinguishing the change point in the historical characteristic data sequence into the first change point having a larger degree of change and the second change point having a smaller degree of change; segmenting the original historical characteristic data sequence based on the detected first change point and the second change point, thereby obtaining a plurality of data segments with higher quality compared with the original historical characteristic data sequence; and training the predictor by using the obtained plurality of data segments, the performance of the predictor can be improved. In the case where the object is various parts involved in an automobile part supply service, theinformation processing method 400 according to the embodiment of the present disclosure can more accurately reflect the degree of change of data in the historical inventory data sequence of automobile parts; segmenting the original inventory data sequence based on the detected first change point and the second change point, thereby obtaining a plurality of data segments with higher quality compared with the original inventory data sequence; and by predicting the inventory requirements of the automotive parts using the plurality of data segments, inventory management can be optimized on a predictive basis.

The present disclosure also provides an information processing apparatus for predicting time-varying characteristic data of an object to be processed. Hereinafter, an information processing apparatus that predicts time-varying characteristic data of an object to be processed is referred to as a prediction apparatus for distinction from theinformation processing apparatus 100. The prediction means comprises a trained predictor obtained by theinformation processing apparatus 100. In the prediction apparatus, the trained predictor may be configured to predict time-varying characteristic data of the object to be processed.

Since, in theinformation processing apparatus 100 according to the embodiment of the present disclosure, the quality of training data is improved to improve the performance of the predictor, the prediction apparatus described above can improve the prediction accuracy.

In correspondence with the above-described embodiment of the information processing apparatus for predicting time-varying characteristic data of an object to be processed, the present disclosure also provides an embodiment of an information processing method.

Hereinafter, in order to distinguish from theinformation processing method 400, an information processing apparatus for predicting time-varying characteristic data of an object to be processed is referred to as a prediction method. In this prediction method, the trained predictor obtained by theinformation processing method 400 is used to predict time-varying characteristic data of the object to be processed.

Since in theinformation processing method 400 according to the embodiment of the present disclosure, the quality of training data is improved to improve the performance of the predictor, the prediction method described above can improve the prediction accuracy.

It should be noted that although the information processing apparatus and method according to the embodiments of the present disclosure are described above, this is merely an example and not a limitation, and a person skilled in the art may modify the above embodiments according to the principles of the present disclosure, for example, functional blocks and operations in the respective embodiments may be added, deleted, or combined, and such modifications fall within the scope of the present disclosure.

In addition, it should be further noted that the method embodiments herein correspond to the apparatus embodiments described above, and therefore, the contents that are not described in detail in the method embodiments may refer to the descriptions of the corresponding parts in the apparatus embodiments, and the description is not repeated here.

In addition, the present disclosure also provides a storage medium and a program product. The machine-executable instructions in the storage medium and the program product according to the embodiments of the present disclosure may be configured to perform the above-described information processing method, and thus, a content not described in detail herein may refer to the description of the corresponding parts previously, and the description will not be repeated herein.

Accordingly, storage media for carrying the above-described program products comprising machine-executable instructions are also included in the present disclosure. Including, but not limited to, floppy disks, optical disks, magneto-optical disks, memory cards, memory sticks, and the like.

Further, it should be noted that the above series of processes and means may also be implemented by software and/or firmware. In the case of implementation by software and/or firmware, a program constituting the software is installed from a storage medium or a network to a computer having a dedicated hardware structure, such as a general-purposepersonal computer 800 shown in fig. 5, which is capable of executing various functions and the like when various programs are installed.

In fig. 5, a Central Processing Unit (CPU)801 executes various processes in accordance with a program stored in a Read Only Memory (ROM)802 or a program loaded from astorage section 808 to a Random Access Memory (RAM) 803. In theRAM 803, data necessary when theCPU 801 executes various processes and the like is also stored as necessary.

TheCPU 801, theROM 802, and theRAM 803 are connected to each other via abus 804. An input/output interface 805 is also connected to thebus 804.

The following components are connected to the input/output interface 805: aninput section 806 including a keyboard, a mouse, and the like; anoutput section 807 including a display such as a Cathode Ray Tube (CRT), a Liquid Crystal Display (LCD), and the like, and a speaker and the like; astorage section 808 including a hard disk and the like; and acommunication section 809 including a network interface card such as a LAN card, a modem, and the like. Thecommunication section 809 performs communication processing via a network such as the internet.

Adrive 810 is also connected to the input/output interface 805 as needed. Aremovable medium 811 such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, or the like is mounted on thedrive 810 as necessary, so that a computer program read out therefrom is installed in thestorage portion 808 as necessary.

In the case where the above-described series of processes is realized by software, a program constituting the software is installed from a network such as the internet or a storage medium such as theremovable medium 811.

It will be understood by those skilled in the art that such a storage medium is not limited to theremovable medium 811 shown in fig. 5 in which the program is stored, distributed separately from the apparatus to provide the program to the user. Examples of theremovable medium 811 include a magnetic disk (including a floppy disk (registered trademark)), an optical disk (including a compact disk read only memory (CD-ROM) and a Digital Versatile Disk (DVD)), a magneto-optical disk (including a Mini Disk (MD) (registered trademark)), and a semiconductor memory. Alternatively, the storage medium may be theROM 802, a hard disk included in thestorage section 808, or the like, in which programs are stored and which are distributed to users together with the apparatus including them.

The preferred embodiments of the present disclosure are described above with reference to the drawings, but the present disclosure is of course not limited to the above examples. Various changes and modifications within the scope of the appended claims may be made by those skilled in the art, and it should be understood that these changes and modifications naturally will fall within the technical scope of the present disclosure.

For example, a plurality of functions included in one unit may be implemented by separate devices in the above embodiments. Alternatively, a plurality of functions implemented by a plurality of units in the above embodiments may be implemented by separate devices, respectively. In addition, one of the above functions may be implemented by a plurality of units. Needless to say, such a configuration is included in the technical scope of the present disclosure.

In this specification, the steps described in the flowcharts include not only the processing performed in time series in the described order but also the processing performed in parallel or individually without necessarily being performed in time series. Further, even in the steps processed in time series, needless to say, the order can be changed as appropriate.

In addition, the technique according to the present disclosure can also be configured as follows.

Supplementary note 1. an information processing apparatus comprising:

a change point detecting unit configured to detect a plurality of change points from a history characteristic data sequence of an object using a first data dividing unit, wherein a first degree of change in a characteristic of a first change point of the plurality of change points with respect to a statistical characteristic of data in the first data dividing unit is greater than a second degree of change in a characteristic of a second change point of the plurality of change points with respect to the statistical characteristic of data in the first data dividing unit;

a historical characteristic data sequence dividing unit configured to divide the historical characteristic data sequence into a plurality of data segments based on the detected first change point and the second change point; and

a training unit configured to train a predictor of the subject based on the plurality of data segments, wherein the predictor is used to predict time-varying characteristic data of the subject.

Supplementary note 2. the information processing apparatus according to supplementary note 1, wherein,

the change point detecting unit is configured to slide the first data dividing unit over the history characteristic data sequence, and detect history characteristic data at a predetermined position outside the first data dividing unit as the first change point or the second change point, in a case where the history characteristic data satisfies a predetermined condition.

Note 3 that the information processing apparatus according to note 2, wherein,

the change point detecting unit is configured to define the predetermined condition with a quartile of the history characteristic data in the first data dividing unit, wherein the first degree of change and the second degree of change are degrees of change in magnitudes of the first change point and the second change point, respectively, with respect to a magnitude determined based on the quartile.

Note 4. the information processing apparatus according to note 1, wherein,

the historical characteristic data sequence dividing unit is configured to search for a dividing point for dividing the historical characteristic data sequence from the detected first change point and second change point, using a second data dividing unit to slide on the historical characteristic data sequence, wherein the dividing point is not included in any of the plurality of data segments.

Supplementary note 5. the information processing apparatus according to supplementary note 4, wherein,

the history characteristic data sequence dividing unit is configured to, when the first change point is found in the second data dividing unit or at an end point of the second data dividing unit, take the found first change point as the dividing point, and move the start point of the second data dividing unit to the found first change point to continue sliding on the history characteristic data sequence.

Supplementary note 6. the information processing apparatus according to supplementary note 4 or 5, wherein,

the history characteristic data sequence dividing unit is configured to, when an end point of the second data dividing unit is the second change point, take the second change point as the end point as the dividing point, and move the start point of the second data dividing unit to the second change point as the end point to continue sliding on the history characteristic data sequence.

Note 7. the information processing apparatus according to note 4 or 5, wherein,

the history characteristic data sequence dividing unit is configured to make the second data dividing unit slide in a direction from the end to the start of the history characteristic data sequence with the end of the history characteristic data sequence as an initial start of the second data dividing unit.

Note 8 that the information processing apparatus according to note 1, wherein,

the training unit is configured to train the predictor separately using each of the plurality of data segments.

Note 9 that the information processing apparatus according to note 1, wherein,

the training unit is configured to cluster the plurality of data segments and train the predictor using the clustered data segments.

Reference numeral 10 denotes the information processing apparatus according to reference numeral 1, wherein the object is various parts relating to an automobile parts supply service.

Note 11. an information processing method includes:

a change point detection step of detecting a plurality of change points from a history characteristic data sequence of an object using a first data dividing unit, wherein a first degree of change in a characteristic of a first change point of the plurality of change points with respect to a statistical characteristic of data in the first data dividing unit is larger than a second degree of change in a characteristic of a second change point of the plurality of change points with respect to a statistical characteristic of data in the first data dividing unit;

a historical characteristic data sequence dividing step of dividing the historical characteristic data sequence into a plurality of data segments based on the detected first change point and second change point; and

training a predictor of the object based on the plurality of data segments, wherein the predictor is used for predicting the characteristic data of the object changing along with time.

Note 12 that the information processing method according to note 11, wherein,

in the change point detecting step, the first data dividing unit is caused to slide on the history characteristic data sequence, and in a case where history characteristic data at a predetermined position outside the first data dividing unit satisfies a predetermined condition, the history characteristic data is detected as the first change point or the second change point.

Note 13. the information processing method according to note 12, wherein,

in the change point detecting step, the predetermined condition is defined by a quartile of the history characteristic data in the first data dividing unit, wherein the first degree of change and the second degree of change are degrees of change in the magnitudes of the first change point and the second change point, respectively, with respect to a magnitude determined based on the quartile.

Supplementary note 14. the information processing method according to supplementary note 11, wherein,

in the historical characteristic data sequence dividing step, a dividing point for dividing the historical characteristic data sequence is searched from the detected first change point and second change point by sliding on the historical characteristic data sequence using a second data dividing unit, wherein the dividing point is not included in any of the plurality of data segments.

Supplementary note 15. the information processing method according to supplementary note 14, wherein,

in the historical characteristic data sequence dividing step, when the first change point is found in the second data dividing unit or at an end point of the second data dividing unit, the found first change point is taken as the dividing point, and the start point of the second data dividing unit is moved to the found first change point to continue sliding on the historical characteristic data sequence.

Note 16. the information processing method according to note 14 or 15, wherein,

in the history characteristic data sequence dividing step, when the end point of the second data dividing unit is the second change point, the second change point as the end point is made the dividing point, and the start point of the second data dividing unit is moved to the second change point as the end point to continue sliding on the history characteristic data sequence.

Supplementary note 17 the information processing method according to supplementary note 14 or 15, wherein,

in the historical characteristic data sequence dividing step, the end of the historical characteristic data sequence is used as the initial starting point of the second data dividing unit, and the second data dividing unit is made to slide in the direction from the end to the starting point of the historical characteristic data sequence.

Note 18. the information processing method according to note 11, wherein,

in the training step, the predictor is trained separately using each of the plurality of data segments.

Supplementary note 19. the information processing method according to supplementary note 11, wherein,

in the training step, the plurality of data segments are clustered, and the predictor is trained using the clustered data segments.

20. An information processing apparatus comprising the trained predictor obtained by the information processing device according to any one of claims 1 to 10, wherein

The trained predictor is configured to predict time-varying characteristic data of an object to be processed.