CN102215300A - Telecommunication service recommendation method and system - Google Patents

Abstract

Translated fromChinese

本发明提供一种电信业务推荐方法和系统，其中方法包括：获取电信业务信息、用户的身份信息、订购业务信息和上网日志信息；根据所述上网日志信息得到用户访问网页文档的主题标引；根据所述用户访问网页文档主题标引和上网日志信息建立用户兴趣模型；根据所述用户兴趣模型、电信业务信息和订购业务信息，获得用户兴趣和电信业务的关联关系；并根据所述用户兴趣和电信业务的关联关系以及所述用户兴趣模型，得到推荐业务列表，对用户进行电信业务推荐。本发明实现了电信业务的精准营销，同时提升用户体验。

The present invention provides a method and system for recommending telecommunication services, wherein the method includes: acquiring telecommunication service information, user identity information, ordering service information, and online log information; obtaining subject indexing of webpage documents accessed by users according to the online log information; Establish a user interest model according to the user's access to web page document subject indexing and online log information; obtain the association between user interest and telecommunication service according to the user interest model, telecommunication service information and subscription service information; and according to the user interest Based on the association relationship with the telecommunications service and the user interest model, a recommended service list is obtained, and the telecommunications service is recommended to the user. The present invention realizes precise marketing of telecommunication services, and improves user experience at the same time.

Description

Translated fromChinese

电信业务推荐方法和系统Telecom service recommendation method and system

技术领域technical field

本发明涉及通信技术，特别涉及一种电信业务推荐方法和系统。The invention relates to communication technology, in particular to a method and system for recommending telecommunication services.

背景技术Background technique

现阶段我国电信业务面临同质化竞争的市场压力，使得发展新用户越来越困难；电信运营商正逐步从以往的依赖价格战等竞争手段吸引用户的粗放式经营，向分众服务、精细化经营转变，纷纷建设自己的数据挖掘系统来辅助进行精准营销。At present, my country's telecommunications business is facing market pressure of homogeneous competition, which makes it more and more difficult to develop new users; telecom operators are gradually shifting from the previous extensive operation of relying on price wars and other competitive means to attract users to focused services and refined services. Due to business changes, they have built their own data mining systems to assist in precision marketing.

目前一般是基于电信用户的基本资料、地理位置、通信行为等数据的分析和挖掘，从中发现业务运营规律，进而开展电信业务推荐。该方法的弊端是，没有考虑用户的兴趣爱好，用户的兴趣爱好无法直接从用户的基本资料、地理位置、通信行为等数据获得，可能出现用户对推荐的电信业务不感兴趣，既使得用户体验较差，又降低了电信业务推荐的成功率。At present, it is generally based on the analysis and mining of basic information, geographical location, communication behavior and other data of telecom users, from which business operation rules are discovered, and then telecom service recommendations are carried out. The disadvantage of this method is that it does not consider the user's hobbies. The user's hobbies cannot be directly obtained from the user's basic information, geographical location, communication behavior and other data. It may appear that the user is not interested in the recommended telecommunication services, which makes the user experience relatively poor. Poor, which reduces the success rate of telecom service recommendation.

发明内容Contents of the invention

本发明的目的是提供一种电信业务推荐方法和系统，以基于用户兴趣进行电信业务推荐，实现电信业务的精准营销。The purpose of the present invention is to provide a method and system for recommending telecommunication services, so as to recommend telecommunication services based on user interests and realize precise marketing of telecommunication services.

本发明提供一种电信业务推荐方法，包括：The present invention provides a method for recommending telecommunication services, including:

获取电信业务信息、用户的身份信息、订购业务信息和上网日志信息；Obtain telecommunications business information, user identity information, subscription business information and online log information;

根据所述上网日志信息得到用户访问网页文档的主题标引；根据所述用户访问网页文档的主题标引和上网日志信息建立用户兴趣模型；Obtain the subject index of the user's access to the webpage document according to the log information on the Internet; establish a user interest model according to the subject index of the user's access to the webpage document and the log information on the Internet;

根据所述用户兴趣模型、电信业务信息和订购业务信息，获得用户兴趣和电信业务的关联关系；并根据所述用户兴趣和电信业务的关联关系以及所述用户兴趣模型，得到推荐业务列表，对用户进行电信业务推荐。According to the user interest model, telecommunication service information and subscription service information, obtain the association relationship between user interest and telecommunication service; and according to the user interest and telecommunication service association relationship and the user interest model, obtain a list of recommended services, for The user recommends telecommunications services.

本发明提供一种电信业务推荐系统，包括：用户数据采集模块、用户兴趣学习模块和电信业务推荐模块；The present invention provides a telecommunication service recommendation system, comprising: a user data collection module, a user interest learning module and a telecommunication service recommendation module;

所述用户数据采集模块，用于获取电信业务信息、用户的身份信息、订购业务信息和上网日志信息；The user data collection module is used to obtain telecommunication service information, user identity information, subscription service information and log information;

所述用户兴趣学习模块，用于根据所述上网日志信息得到用户访问网页文档的主题标引；根据所述用户访问网页文档的主题标引和上网日志信息建立用户兴趣模型；The user interest learning module is used to obtain the subject index of the user's access to the webpage document according to the log information on the Internet; establish a user interest model according to the subject index of the user's access to the webpage document and the log information on the Internet;

所述电信业务推荐模块，用于根据所述用户兴趣模型、电信业务信息和订购业务信息，获得用户兴趣和电信业务的关联关系；并根据所述用户兴趣和电信业务的关联关系以及所述用户兴趣模型，得到推荐业务列表，对用户进行电信业务推荐。The telecommunications service recommendation module is configured to obtain the association between user interests and telecommunications services according to the user interest model, telecommunications service information, and subscription service information; Interest model, get the list of recommended services, and recommend telecom services to users.

本发明的电信业务推荐方法和系统，通过对全业务运营环境下电信用户的上网行为的采集和分析，充分挖掘用户兴趣及其与电信业务之间的关联关系，基于用户兴趣开展电信业务推荐，实现电信业务的精准营销，同时提升用户体验。The telecommunications service recommendation method and system of the present invention, through the collection and analysis of online behaviors of telecommunications users in a full-service operation environment, fully excavates user interests and their association with telecommunications services, and carries out telecommunications service recommendations based on user interests, Realize the precise marketing of telecom services while improving user experience.

附图说明Description of drawings

为了更清楚地说明本发明实施例或现有技术中的技术方案，下面将对实施例或现有技术描述中所需要使用的附图作一简单地介绍，显而易见地，下面描述中的附图是本发明的一些实施例，对于本领域普通技术人员来讲，在不付出创造性劳动的前提下，还可以根据这些附图获得其他的附图。In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings that need to be used in the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description These are some embodiments of the present invention. Those skilled in the art can also obtain other drawings based on these drawings without creative work.

图1为本发明电信业务推荐系统实施例的结构示意图；Fig. 1 is a schematic structural diagram of an embodiment of the telecommunication service recommendation system of the present invention;

图2为本发明电信业务推荐方法实施例的流程示意图；FIG. 2 is a schematic flow diagram of an embodiment of a method for recommending telecommunication services according to the present invention;

图3为本发明电信业务推荐方法实施例中的生成文本分类器的流程示意图；FIG. 3 is a schematic flow diagram of generating a text classifier in an embodiment of the telecommunications service recommendation method of the present invention;

图4为本发明电信业务推荐方法实施例中的获得用户访问网页文档主题标引的流程示意图；4 is a schematic flow diagram of obtaining subject indexing of webpage documents accessed by users in an embodiment of the telecommunications service recommendation method of the present invention;

图5为本发明电信业务推荐方法实施例中的获得用户兴趣和电信业务关联关系的流程示意图；FIG. 5 is a schematic flow diagram of obtaining user interests and telecommunication service associations in an embodiment of the telecommunication service recommendation method of the present invention;

图6为本发明电信业务推荐方法实施例中的电信业务推荐的流程示意图。FIG. 6 is a schematic flow chart of telecommunications service recommendation in an embodiment of the telecommunications service recommendation method of the present invention.

具体实施方式Detailed ways

为使本发明的目的、技术方案和优点更加清楚，下面将结合本发明实施例中的附图，对本发明实施例中的技术方案进行清楚、完整地描述，显然，所描述的实施例是本发明一部分实施例，而不是全部的实施例。基于本发明中的实施例，本领域普通技术人员在没有做出创造性劳动的前提下所获得的所有其他实施例，都属于本发明保护的范围。In order to make the purpose, technical solutions and advantages of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are the Some, but not all, embodiments are invented. Based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without making creative efforts belong to the protection scope of the present invention.

实施例一Embodiment one

图1为本发明电信业务推荐系统实施例的结构示意图，如图1所示，该系统可以包括用户数据采集模块11、用户兴趣学习模块12和电信业务推荐模块13；Fig. 1 is a schematic structural diagram of an embodiment of a telecommunication service recommendation system of the present invention, as shown in Fig. 1 , the system may include a userdata collection module 11, a userinterest learning module 12 and a telecommunicationservice recommendation module 13;

其中，所述用户数据采集模块11，用于获取电信业务信息、用户的身份信息、订购业务信息和上网日志信息；该上网日志信息可以存储至用户上网日志库14。Wherein, the userdata acquisition module 11 is used to acquire telecommunication service information, user identity information, subscription service information and online log information; the online log information can be stored in the useronline log library 14 .

所述用户兴趣学习模块12，用于根据所述上网日志信息得到用户访问网页文档主题标引；根据所述用户访问网页文档主题标引和上网日志信息建立用户兴趣模型；该用户兴趣模型可以存储至用户兴趣模型库15；The userinterest learning module 12 is used to obtain the user access webpage document subject index according to the online log information; set up a user interest model according to the user access web document subject index and online log information; the user interest model can be stored To the userinterest model library 15;

所述电信业务推荐模块13，用于根据所述用户兴趣模型、电信业务信息和订购业务信息，获得用户兴趣和电信业务的关联关系；并根据所述用户兴趣和电信业务的关联关系以及所述用户兴趣模型，得到推荐业务列表，对用户进行电信业务推荐；该用户兴趣和电信业务的关联关系可以存储至兴趣-业务关联库16。The telecommunicationsservice recommending module 13 is configured to obtain an association between user interests and telecommunications services according to the user interest model, telecommunications service information, and subscription service information; The user interest model obtains the list of recommended services and recommends telecommunication services to users; the relationship between user interests and telecommunication services can be stored in the interest-service association library 16 .

例如，用户兴趣学习模块12可以包括判断单元、第一处理单元和第二处理单元；其中，判断单元，用于根据所述上网日志信息得到用户所访问的URL地址，并根据所述URL地址查询URL-主题映射库，判断该URL地址对应的用户访问网页文档是否已经进行主题标引；第一处理单元，用于在所述判断单元的判断结果为是时，直接获得所述主题标引；第二处理单元，用于在所述判断单元的判断结果为否时，使用网页采集器获取所述URL对应的用户访问网页文档，并对所述用户访问网页文档进行预处理，得到规范化文本；对所述规范化文本进行中文分词和英文词组辨认，去除非分类信息词，得到带分词标识文本；对所述带分词标识文本进行词频统计和特征权重计算，生成文本向量；并对所述文本向量进行文本分类，得到所述用户访问网页文档主题标引；将所述用户访问网页文档主题标引存储到所述URL-主题映射库。For example, the userinterest learning module 12 may include a judging unit, a first processing unit, and a second processing unit; wherein, the judging unit is used to obtain the URL address visited by the user according to the log information on the Internet, and query according to the URL address URL-theme mapping library, judging whether the user's access webpage document corresponding to the URL address has been subject indexed; the first processing unit is used to directly obtain the subject index when the judgment result of the judging unit is yes; The second processing unit is configured to use a web page collector to acquire a user-accessed web page document corresponding to the URL when the determination result of the determining unit is negative, and preprocess the user-accessed web page document to obtain a standardized text; Carry out Chinese word segmentation and English phrase recognition to described standardization text, remove non-categorization information words, obtain band segmentation identification text; Carry out word frequency statistics and feature weight calculation to described band segmentation identification text, generate text vector; And described text vector Perform text classification to obtain the subject index of the web page document accessed by the user; store the subject index of the web page document accessed by the user in the URL-subject mapping library.

例如，用户兴趣学习模块12还可以包括第一兴趣单元和第二兴趣单元；其中，第一兴趣单元，用于根据所述用户访问网页文档主题标引和上网日志信息建立用户兴趣模型，所述用户兴趣模型中包括多个用户兴趣项，所述用户兴趣项中包括兴趣类别、与所述兴趣类别对应的兴趣度；所述用户兴趣的数量等于所述用户访问网页文档主题标引的数量，所述兴趣度与所述上网日志信息中的与所述兴趣类别对应的日志数量成正比，与所述上网日志信息中的与所述兴趣类别对应的日志数量的最近出现时间成反比；第二兴趣单元，用于对所述兴趣度进行正规化处理，得到正规兴趣度。For example, the userinterest learning module 12 can also include a first interest unit and a second interest unit; wherein, the first interest unit is used to set up a user interest model according to the subject index of the user's access to webpage documents and log information on the Internet, the The user interest model includes a plurality of user interest items, and the user interest item includes an interest category and an interest degree corresponding to the interest category; the quantity of the user interest is equal to the quantity of subject indexing of the web page document accessed by the user, The degree of interest is directly proportional to the number of logs corresponding to the interest category in the online log information, and inversely proportional to the latest appearance time of the log quantity corresponding to the interest category in the online log information; the second The interest unit is configured to normalize the interest degree to obtain a regular interest degree.

例如，所述电信业务推荐模块包括：集合单元、处理单元、关联单元、可信单元和推荐单元；For example, the telecommunications service recommendation module includes: an aggregation unit, a processing unit, an association unit, a credible unit and a recommendation unit;

集合单元，用于根据所述用户兴趣模型、电信业务信息和订购业务信息，建立二维带权记录集合；An aggregation unit, configured to establish a two-dimensional weighted record set according to the user interest model, telecommunication service information and subscription service information;

处理单元，用于计算所述集合中各类兴趣的支持度；计算所述集合中的兴趣和业务二维项集的支持度；A processing unit, configured to calculate the support degree of various interests in the set; calculate the support degree of the interest and business two-dimensional item sets in the set;

关联单元，用于根据所述各类兴趣的支持度以及二维项集的支持度，得到用户兴趣和电信业务的关联规则的置信度；An associating unit, configured to obtain the confidence degree of the user interest and the association rule of the telecommunication service according to the support degree of the various types of interests and the support degree of the two-dimensional item set;

可信单元，用于根据所述用户兴趣模型、以及所述用户兴趣和电信业务的关联规则的置信度，得到电信业务推荐的可信度；a credible unit, configured to obtain the credibility of the telecommunications service recommendation according to the user interest model and the confidence of the user interest and the association rules of the telecommunications service;

推荐单元，用于根据所述电信业务推荐的可信度，将所待推荐业务排列，生成所述推荐业务列表。The recommendation unit is configured to arrange the services to be recommended according to the credibility of the telecommunication service recommendation, and generate the recommended service list.

进一步的，还包括：推荐评估模块17，用于对电信业务推荐的效果进行评估，得到评估结果；推荐优化模块18，用于根据所述评估结果，优化所述电信业务推荐。Further, it also includes: arecommendation evaluation module 17, configured to evaluate the effect of the telecommunication service recommendation, and obtain an evaluation result; arecommendation optimization module 18, configured to optimize the telecommunication service recommendation according to the evaluation result.

本实施例的电信业务推荐方法，通过根据用户的上网日志信息建立用户兴趣模型，以及用户兴趣和电信业务的关联关系等，可以据此生成推荐业务列表，对用户进行电信业务推荐，解决了电信业务推荐的成功率低的问题，实现了基于用户兴趣进行电信业务推荐，以及电信业务的精准营销，提升电信业务推荐的用户体验。The method for recommending telecommunications services in this embodiment establishes a user interest model based on the user's log information on the Internet, as well as the association between user interests and telecommunications services, etc., so as to generate a list of recommended services and recommend telecommunications services to users, solving the problem of telecommunications. The problem of the low success rate of service recommendation enables telecom service recommendation based on user interests and precise marketing of telecom services to improve the user experience of telecom service recommendation.

实施例二Embodiment two

图2为本发明电信业务推荐方法实施例的流程示意图，本实施例的电信业务推荐方法可以通过实施例一中所述的电信业务推荐系统执行，如图2所示，本实施例的方法可以包括以下步骤：Fig. 2 is a schematic flow chart of an embodiment of the method for recommending telecommunication services in the present invention. The method for recommending telecommunication services in this embodiment can be implemented by the system for recommending telecommunication services described in Embodiment 1. As shown in Fig. 2, the method in this embodiment can be Include the following steps:

步骤201、用户数据采集模块获取电信业务信息、用户的身份信息、订购业务信息和上网日志信息；Step 201, the user data acquisition module acquires telecommunication service information, user identity information, subscription service information and online log information;

例如，可以由BSS/OSS系统采集用户身份信息、订购业务信息、电信业务信息等信息，由宽带接入服务器(Broadband Remote Access Server，简称：BRAS)和远程用户拨号认证系统(Remote Authentication Dial In User Service，简称：RADIUS)采集用户访问互联网的原始信息，得到用户访问URL的上网日志信息；该上网行为包括固网宽带、移动WAP/Web上网等。所述的上网日志信息可以存储到用户上网日志库。For example, the BSS/OSS system can collect information such as user identity information, subscription service information, and telecommunication service information, and the broadband access server (Broadband Remote Access Server, referred to as: BRAS) and remote user dial-in authentication system (Remote Authentication Dial In User Service, referred to as: RADIUS) collects the original information of the user's access to the Internet, and obtains the log information of the user's access to the URL; the Internet access behavior includes fixed network broadband, mobile WAP/Web access, etc. The online log information may be stored in a user online log library.

其中，用户的身份信息用于用户数据采集模块从BRAS和RADIUS采集用户使用互联网的原始信息后的用户身份鉴别，以及用户爱好(由上网日志反映)和用户订购业务信息的关联关系挖掘；订购业务信息中可以包括订购业务数量、订购电信业务名称等；电信业务信息中可以包括电信业务总数、电信业务名称等；上网日志信息可以包括用户访问的URL地址信息、某类日志的数量、以及访问该类日志的时间等。Among them, the user's identity information is used for user identity identification after the user data acquisition module collects the original information of the user's use of the Internet from BRAS and RADIUS, and mining of the association relationship between the user's preferences (reflected by the log on the Internet) and the user's order service information; order service The information may include the number of subscribed services, the name of the ordered telecommunication service, etc.; the telecommunication service information may include the total number of telecommunication services, the name of the telecommunication service, etc.; the log information of the Internet may include the URL address information accessed by the user, the number of certain types of logs, and the class log time etc.

步骤202、用户兴趣学习模块根据所述上网日志信息得到用户访问网页文档的主题标引；Step 202, the user interest learning module obtains the subject index of the webpage document accessed by the user according to the online log information;

例如，可以由上网日志信息中获得用户访问的URL地址，根据所述上网日志信息得到用户所访问的URL地址，并根据所述URL地址查询URL-主题映射库，判断该URL地址对应的用户访问网页文档主题是否已标引。在所述判断单元的判断结果为是时，直接获得所述主题标引。在所述判断单元的判断结果为否时，使用网页采集器，采集获得所述URL地址所对应的网页文档，即用户访问网页文档；对所述用户访问网页文档进行处理，得到用户访问网页文档主题标引；并将所述用户访问网页文档主题标引存储到所述URL-主题映射库。For example, the URL address accessed by the user can be obtained from the log information on the Internet, the URL address accessed by the user can be obtained according to the log information on the Internet, and the URL-topic mapping library can be queried according to the URL address to determine the user access corresponding to the URL address. Whether the web document topic is indexed. When the judging result of the judging unit is yes, the subject index is obtained directly. When the judgment result of the judging unit is no, use the webpage collector to collect and obtain the webpage document corresponding to the URL address, that is, the user accesses the webpage document; process the user's access to the webpage document to obtain the user's access to the webpage document Subject indexing; and storing the subject index of the web page document accessed by the user in the URL-subject mapping library.

其中，对所述用户访问网页文档进行处理，得到用户访问网页文档主题标引，具体可以包括两部分内容；一部分是可以通过对有主题类别标注的分类训练文档库的文档进行文本分类训练，生成文本分类器，该部分内容具体可以参见实施例三中对文本分类训练的说明；另一部分是在对用户访问网页文档的处理过程中使用上述得到的文本分类器进行文本分类，从而得到用户访问网页文档的主题标引，该部分内容具体可以参见实施例四中获得用户访问网页文档主题标引的说明。Wherein, the web page document accessed by the user is processed to obtain the subject index of the web page document accessed by the user, which may specifically include two parts; one part is to perform text classification training on the documents of the classification training document library marked with subject categories to generate Text classifier, this part of the content can refer to the description of text classification training in embodiment three; the other part is to use the text classifier obtained above to classify the text in the process of user access to web page documents, thereby obtaining the user access web page For the subject index of the document, please refer to the description of obtaining the subject index of the web page document accessed by the user in the fourth embodiment.

步骤203、用户兴趣学习模块根据所述用户访问网页文档主题标引和上网日志信息建立用户兴趣模型；Step 203, the user interest learning module establishes a user interest model according to the user's access to webpage document subject indexing and online log information;

例如，可以根据用户访问网页文档主题标引确定用户的兴趣数量，该用户兴趣模型中，可以包括多个兴趣项，其中，所述的兴趣项中可以包括兴趣类别名称和兴趣度等因素。该用户兴趣模型具体可以参见实施例五中对于用户兴趣模型的详细说明。生成的用户兴趣模型可以存储至用户兴趣模型库。For example, the user's interest quantity can be determined according to the subject index of the webpage document visited by the user, and the user interest model can include multiple interest items, wherein the interest items can include factors such as interest category names and interest degrees. For details of the user interest model, refer to the detailed description of the user interest model in Embodiment 5. The generated user interest model can be stored in a user interest model library.

步骤204、电信业务推荐模块根据所述用户兴趣模型、电信业务信息和订购业务信息，获得用户兴趣和电信业务的关联关系；Step 204, the telecommunication service recommendation module obtains the correlation between user interest and telecommunication service according to the user interest model, telecommunication service information and subscription service information;

例如，可以根据步骤203中所建立的用户兴趣模型，以及步骤201中所获取的电信业务信息和订购业务信息等，通过计算和处理得到用户兴趣和电信业务之间的关联关系。该关联关系的生成具体可以参见实施例六的说明。生成的兴趣-业务关联关系可以存储至兴趣-业务关联库中。For example, according to the user interest model established instep 203, and the telecommunication service information and subscription service information acquired instep 201, the correlation between user interests and telecommunication services can be obtained through calculation and processing. For the generation of the association relationship, please refer to the description of Embodiment 6 for details. The generated interest-business association relationship can be stored in an interest-business association library.

步骤205、电信业务推荐模块根据所述用户兴趣和电信业务的关联关系以及所述用户兴趣模型，得到推荐业务列表，对用户进行电信业务推荐。Step 205, the telecommunications service recommendation module obtains a list of recommended services according to the association between the user's interest and the telecommunications service and the user interest model, and recommends the telecommunications service to the user.

例如，在步骤204中得到用户兴趣和电信业务的关联关系之后，可以根据用户兴趣模型以及所述关联关系，得到推荐业务列表，对用户进行电信业务推荐。具体可以参见实施例七中对于电信业务推荐列表生成的说明。For example, after obtaining the correlation between the user's interest and the telecommunication service instep 204, a list of recommended services can be obtained according to the user interest model and the correlation, and the telecommunication service can be recommended to the user. For details, please refer to the description of the generation of the telecommunication service recommendation list in Embodiment 7.

本实施例的电信业务推荐方法，通过根据用户的上网日志信息建立用户兴趣模型，以及用户兴趣和电信业务的关联关系等，可以据此生成推荐业务列表，对用户进行电信业务推荐，解决了电信业务推荐的成功率低的问题，实现了基于用户兴趣进行电信业务推荐，以及电信业务的精准营销，提升电信业务推荐的用户体验。The method for recommending telecommunications services in this embodiment establishes a user interest model based on the user's log information on the Internet, as well as the association between user interests and telecommunications services, etc., so as to generate a list of recommended services and recommend telecommunications services to users, solving the problem of telecommunications. The problem of the low success rate of service recommendation enables telecom service recommendation based on user interests and precise marketing of telecom services to improve the user experience of telecom service recommendation.

实施例三Embodiment three

图3为本发明电信业务推荐方法实施例中的生成文本分类器的流程示意图，如图3所示，本实施例可以包括以下步骤：Fig. 3 is a schematic flow chart of generating a text classifier in an embodiment of the telecommunications service recommendation method of the present invention. As shown in Fig. 3, this embodiment may include the following steps:

步骤301、用户兴趣学习模块对有类别标注的分类训练文档库的文档进行预处理，得到规范化文本；Step 301, the user interest learning module preprocesses the documents in the classified training document library marked with categories to obtain standardized texts;

例如，用户兴趣学习模块对有主题类别标注的分类训练文档库的文档进行预处理。其中，所述分类训练文档库可以由电信业务推荐系统的使用者自行建立，形式可以是一个名为“示例文档库”的大文件夹和一个名为“示例文档主题标引对照表”的文档或数据表。所述示例文档库内是一些示例文档，所述示例文档主题标引对照表说明各示例文档的主题标引，如“体育”、“音乐”、“财经”等等。根据电信业务推荐的需要，文档主题的颗粒度可以再细化，如“体育”可以再细分为“足球”、“排球”等等。所述示例文档的格式不限，可以是txt纯文本、web文本、word文本、ppt文本、pdf文本等等，txt和web类型以外的文档预处理时还需要做文档识别工作，因此建议系统使用者自行建立文档库时最好选择txt和web类型的示例文档。所述示例文档的语言不限，可以是中文文本、英文文本、中英文混合文本。For example, the user interest learning module preprocesses the documents in the classification training document library marked with subject categories. Wherein, the classification training document library can be established by users of the telecommunication service recommendation system, in the form of a large folder named "Example Document Library" and a document named "Example Document Subject Index Comparison Table" or a data table. There are some sample documents in the sample document library, and the sample document subject index comparison table shows the subject index of each sample document, such as "sports", "music", "finance and economics" and so on. According to the needs of telecom service recommendation, the granularity of document topics can be further refined, for example, "sports" can be subdivided into "soccer", "volleyball" and so on. The format of the sample document is not limited, it can be txt plain text, web text, word text, ppt text, pdf text, etc. Documents other than txt and web types need to be identified during preprocessing, so it is recommended that the system use When building a document library by oneself, it is better to choose sample documents of txt and web types. The language of the example document is not limited, and may be Chinese text, English text, or mixed Chinese and English text.

在预处理时可以去除所述示例文本中的噪声，如广告条、导航条、书签等；并使用词法规则对英文文本进行词法分析，把变形后的词恢复原型，包括名词的复数去除、动词的时态转换、动词第三人称转换、词干抽取等，得到规范化的文本。During preprocessing, the noise in the sample text can be removed, such as advertisement strips, navigation bars, bookmarks, etc.; and the English text can be lexically analyzed using lexical rules, and the deformed words can be restored to their prototypes, including plural removal of nouns, verbs tense conversion, verb third-person conversion, word stem extraction, etc., to obtain standardized text.

步骤302、用户兴趣学习模块对所述规范化文本进行中文分词和英文词组辨认，去除非分类信息词，得到带分词标识文本；Step 302, the user interest learning module performs Chinese word segmentation and English phrase recognition on the standardized text, removes non-classified information words, and obtains a text with word segmentation;

例如，用户兴趣学习模块可以根据分词词典，对步骤301中得到的规范化文本进行中文分词和英文词组辨认，并根据停用词词典，去掉该规范化文本中的非分类信息词，即不含有分类信息的词，例如介词、带从句的副词、冠词、连词、代词、情态动词或语气助词等虚词，得到带分词标识文本。For example, the user interest learning module can perform Chinese word segmentation and English phrase recognition on the standardized text obtained instep 301 according to the word segmentation dictionary, and remove non-categorical information words in the normalized text according to the stop word dictionary, that is, it does not contain classification information Words, such as prepositions, adverbs with clauses, articles, conjunctions, pronouns, modal verbs or modal particles and other function words, get the participle-marked text.

其中，分词就是使用划分符号，对文本进行以词或词组为单位的划分。划分符号可以是形如“\w”的标识符号。例如，“云计算(Cloud Computing)是分布式处理(Distributed Computing)、并行处理(Parallel Computing)和网格计算(Grid Computing)的发展，或者说是这些计算机科学概念的商业实现。”采用“\w”符号分词并去除非分类信息词后可得到：“云计算\wCloud Computing\w分布式处理\wDistributed Computing\w并行处理\w Parallel Computing\w网格计算\w Grid Computing\w发展\w计算机\w科学\w概念\w商业\w实现\w”。Among them, word segmentation is to use division symbols to divide the text into units of words or phrases. The delimiter can be an identifier in the form of "\w". For example, "Cloud Computing is the development of Distributed Computing, Parallel Computing, and Grid Computing, or the commercial implementation of these computer science concepts." Using "\ After word segmentation with w" symbol and removal of non-categorical information words, it can be obtained: "Cloud Computing\wCloud Computing\wDistributed Processing\wDistributed Computing\wParallel Processing\w Parallel Computing\wGrid Computing\w Grid Computing\wDevelopment\w Computer\wscience\wconcept\wbusiness\wimplementation\w".

步骤303、用户兴趣学习模块对带分词标识文本进行词频统计和特征抽取，生成文本向量；Step 303, the user interest learning module performs word frequency statistics and feature extraction on the text with word segmentation to generate a text vector;

例如，用户兴趣学习模块对步骤302得到的带分词标识的文本进行词频统计和特征抽取，得到特征向量空间，由此生成文本的向量空间模型，即生成文本向量。For example, the user interest learning module performs word frequency statistics and feature extraction on the text with word segmentation tags obtained instep 302 to obtain a feature vector space, thereby generating a vector space model of the text, that is, generating a text vector.

步骤304、用户兴趣学习模块对文本向量进行分类训练，获得文本分类时所需的参数，即生成文本分类器。Step 304 , the user interest learning module performs classification training on the text vector, and obtains parameters required for text classification, that is, generates a text classifier.

本实施例的电信业务推荐方法，通过根据用户的上网日志信息建立用户兴趣模型，以及用户兴趣和电信业务的关联关系等，可以据此生成推荐业务列表，对用户进行电信业务推荐，解决了电信业务推荐的成功率低的问题，实现了基于用户兴趣进行电信业务推荐，以及电信业务的精准营销，提升电信业务推荐的用户体验。The method for recommending telecommunications services in this embodiment establishes a user interest model based on the user's log information on the Internet, as well as the association between user interests and telecommunications services, etc., so as to generate a list of recommended services and recommend telecommunications services to users, solving the problem of telecommunications. The problem of the low success rate of service recommendation enables telecom service recommendation based on user interests and precise marketing of telecom services to improve the user experience of telecom service recommendation.

实施例四Embodiment Four

图4为本发明电信业务推荐方法实施例中的获得用户访问网页文档主题标引的流程示意图，如图4所示，本实施例可以包括以下步骤：Fig. 4 is a schematic flow diagram of obtaining the subject index of a web page document accessed by a user in an embodiment of the telecommunications service recommendation method of the present invention. As shown in Fig. 4, this embodiment may include the following steps:

步骤401、用户兴趣学习模块获取用户访问的URL地址；Step 401, the user interest learning module obtains the URL address accessed by the user;

例如，用户兴趣学习模块可以由用户上网日志库获取用户访问日志，其中包括URL地址信息。For example, the user interest learning module may obtain user access logs from the user surfing log library, including URL address information.

步骤402、用户兴趣学习模块判断该URL地址所对应的网页文档主题是否已经标引；Step 402, the user interest learning module judges whether the subject of the webpage document corresponding to the URL address has been indexed;

例如，用户兴趣学习模块可以查询URL-主题映射库，判断该URL地址所对应的网页文档主题是否已经标引；如果用户访问的网页文档主题已经标引，则执行步骤403；否则，如果用户访问的网页文档主题未标引，则执行步骤404；For example, the user interest learning module can query the URL-topic mapping library to determine whether the subject of the webpage document corresponding to the URL address has been indexed; if the subject of the webpage document accessed by the user has been indexed, then performstep 403; The subject of the webpage document is not indexed, then step 404 is executed;

步骤403、用户兴趣学习模块获得用户访问网页文档主题标引；Step 403, the user interest learning module obtains the subject index of the webpage document visited by the user;

步骤404、用户兴趣学习模块使用网页采集器，获得步骤401中所获得的URL地址对应用户访问网页文档；Step 404, the user interest learning module uses the webpage collector to obtain the URL address obtained instep 401 corresponding to the user's access to the webpage document;

步骤405、用户兴趣学习模块对用户访问网页文档进行预处理，得到规范化文本；Step 405, the user interest learning module preprocesses the webpage document accessed by the user to obtain a standardized text;

例如，用户兴趣学习模块对步骤404中得到的用户访问网页文档进行预处理，该文档可以为web文本，在预处理时可以去除web文本中的噪声，如广告条、导航条、书签等；并使用词法规则对英文文本进行词法分析，把变形后的词恢复原型，包括名词的复数去除、动词的时态转换、动词第三人称转换、词干抽取等，得到规范化的文本。For example, the user interest learning module preprocesses the web page document accessed by the user obtained instep 404. The document can be web text, and noise in the web text can be removed during preprocessing, such as advertisement bars, navigation bars, bookmarks, etc.; and Use lexical rules to perform lexical analysis on the English text, and restore the deformed words to their prototypes, including plural removal of nouns, tense conversion of verbs, third-person conversion of verbs, stem extraction, etc., to obtain standardized texts.

步骤406、用户兴趣学习模块对所述规范化文本进行中文分词和英文词组辨认，去除非分类信息词，得到带分词标识文本；Step 406, the user interest learning module performs Chinese word segmentation and English phrase recognition on the standardized text, removes non-classified information words, and obtains a text with word segmentation;

例如，用户兴趣学习模块可以根据分词词典，对步骤405中得到的规范化文本进行中文分词和英文词组辨认，并根据停用词词典，去掉该规范化文本中的非分类信息词，即不含有分类信息的词，例如介词、带从句的副词、冠词、连词、代词、情态动词或语气助词等虚词，得到带分词标识文本。For example, the user interest learning module can perform Chinese word segmentation and English phrase recognition on the standardized text obtained instep 405 according to the word segmentation dictionary, and remove non-categorical information words in the normalized text according to the stop word dictionary, that is, it does not contain classification information Words, such as prepositions, adverbs with clauses, articles, conjunctions, pronouns, modal verbs or modal particles and other function words, get the participle-marked text.

步骤407、用户兴趣学习模块对带分词标识文本进行词频统计和特征抽取，生成文本向量；Step 407, the user interest learning module performs word frequency statistics and feature extraction on the text with word segmentation to generate a text vector;

例如，用户兴趣学习模块根据实施例三中得到的特征向量空间，对步骤406得到的带分词标识的文本进行词频统计和特征权重计算，生成文本的向量空间模型，即生成文本向量。For example, the user interest learning module performs word frequency statistics and feature weight calculation on the text with word segmentation tags obtained instep 406 according to the feature vector space obtained in Embodiment 3, and generates a vector space model of the text, that is, generates a text vector.

步骤408、用户兴趣学习模块使用实施例三中得到的文本分类器，对步骤407中得到的文本向量进行文本分类，得到用户访问网页文档主题标引，并存储到URL-主题映射库中。Step 408, the user interest learning module uses the text classifier obtained in the third embodiment to perform text classification on the text vector obtained in step 407, obtains the subject index of the web page document accessed by the user, and stores it in the URL-subject mapping library.

其中，所述主题标引可用于生成用户兴趣模型。所述URL-主题映射库的作用是提高电信业务推荐系统的效率，例如，多个用户都浏览了同一个网页(URL地址相同)，那么电信业务推荐系统只需要在用户首次访问时对这个URL地址指向的网页文档处理一次，后续其他用户再次访问时就可以直接通过查询该映射库得到网页文档对应的主题标引。Wherein, the subject indexing can be used to generate a user interest model. The effect of described URL-topic mapping storehouse is to improve the efficiency of the telecommunication service recommendation system, for example, a plurality of users have browsed same webpage (URL address is identical), so the telecommunication service recommendation system only needs to this URL when the user visits for the first time The webpage document pointed to by the address is processed once, and other users can directly query the mapping library to obtain the subject index corresponding to the webpage document when they visit it again.

本实施例的电信业务推荐方法，通过根据用户的上网日志信息建立用户兴趣模型，以及用户兴趣和电信业务的关联关系等，可以据此生成推荐业务列表，对用户进行电信业务推荐，解决了电信业务推荐的成功率低的问题，实现了基于用户兴趣进行电信业务推荐，以及电信业务的精准营销，提升电信业务推荐的用户体验。The method for recommending telecommunications services in this embodiment establishes a user interest model based on the user's log information on the Internet, as well as the association between user interests and telecommunications services, etc., so as to generate a list of recommended services and recommend telecommunications services to users, solving the problem of telecommunications. The problem of the low success rate of service recommendation enables telecom service recommendation based on user interests and precise marketing of telecom services to improve the user experience of telecom service recommendation.

实施例五Embodiment five

本实施例对实施例二中所述的用户兴趣模型进行详细说明：用户兴趣模型

定义如下式(1)所示：This embodiment describes in detail the user interest model described in Embodiment 2: user interest model

It is defined as shown in the following formula (1):

$\stackrel{<span><span>\&RightArrow;</span>\&Right\; Arrow;</span>}{\mathrm{<span><span>UserInterest</span>UserInterest</span>}}<span><span>=</span>=</span><span><span>\{</span>\{</span><span><span>(</span>(</span>{\mathrm{<span><span>c</span>c</span>}}_{\mathrm{<span><span>0</span>0</span>}}<span><span>,</span>,</span>{\mathrm{<span><span>w</span>w</span>}}_{\mathrm{<span><span>0</span>0</span>}}<span><span>,</span>,</span>{\mathrm{<span><span>n</span>no</span>}}_{\mathrm{<span><span>0</span>0</span>}}<span><span>,</span>,</span>{\mathrm{<span><span>t</span>t</span>}}_{\mathrm{<span><span>0</span>0</span>}}<span><span>)</span>)</span><span><span>,</span>,</span><span><span>\&CenterDot;</span>\&CenterDot;</span><span><span>\&CenterDot;</span>\&CenterDot;</span><span><span>\&CenterDot;</span>\&CenterDot;</span><span><span>,</span>,</span><span><span>(</span>(</span>{\mathrm{<span><span>c</span>c</span>}}_{\mathrm{<span><span>i</span>i</span>}}<span><span>,</span>,</span>{\mathrm{<span><span>w</span>w</span>}}_{\mathrm{<span><span>i</span>i</span>}}<span><span>,</span>,</span>{\mathrm{<span><span>n</span>no</span>}}_{\mathrm{<span><span>i</span>i</span>}}<span><span>,</span>,</span>{\mathrm{<span><span>t</span>t</span>}}_{\mathrm{<span><span>i</span>i</span>}}<span><span>)</span>)</span><span><span>,</span>,</span><span><span>\&CenterDot;</span>\&CenterDot;</span><span><span>\&CenterDot;</span>\&CenterDot;</span><span><span>\&CenterDot;</span>\&Center\; Dot;</span><span><span>,</span>,</span><span><span>(</span>(</span>{\mathrm{<span><span>c</span>c</span>}}_{\mathrm{<span><span>m</span>m</span>}<span><span>-</span>-</span>\mathrm{<span><span>1</span>1</span>}}<span><span>,</span>,</span>{\mathrm{<span><span>w</span>w</span>}}_{\mathrm{<span><span>m</span>m</span>}<span><span>-</span>-</span>\mathrm{<span><span>1</span>1</span>}}<span><span>,</span>,</span>{{\mathrm{<span><span>n</span>no</span>}}_{\mathrm{<span><span>m</span>m</span>}<span><span>-</span>-</span>\mathrm{<span><span>1</span>1</span>}}<span><span>,</span>,</span>\mathrm{<span><span>t</span>t</span>}}_{\mathrm{<span><span>m</span>m</span>}<span><span>-</span>-</span>\mathrm{<span><span>i</span>i</span>}}<span><span>)</span>)</span>\text{<span><span>|0\&le;i\&le;m-1}---</span>|0\&le;i\&le;m-1}---</span>}<span><span>(</span>(</span>\mathrm{<span><span>1</span>1</span>}<span><span>)</span>)</span>$

其中，m(0＜m≤|C|)是该用户的兴趣数量(即用户访问网页文档主题标引的数量)，|C|是主题标引类别总数，(c_i，w_i，n_i，t_i)是该用户的第i+1(0≤i＜m)类兴趣项，c_i是兴趣类别名称(即主题标引)，w_i是c_i类兴趣的权重(兴趣度)，n_i是该用户上网访问日志中c_i类日志的数量，t_i是该用户上网访问日志中c_i类日志的最近出现时间。Among them, m (0<m≤|C|) is the number of interests of the user (that is, the number of subject indexing of webpage documents accessed by the user), |C| is the total number of subject indexing categories, (c_i , w_i , n_i , t_i ) is the i+1 (0≤i<m) category interest item of the user,_ci is the name of the category of interest (that is, subject indexing), w_i is the weight (degree of interest) of the category_ci of interest, n_i is the number of logs of category_ci in the user's online access log, and t_i is the latest appearance time of logs of category_ci in the user's online access log.

该用户的第i+1(0≤i＜m)类兴趣的权重(兴趣度)w_i定义如式(2)所示：The weight (degree of interest) w_i of the user's i+1th (0≤i<m) type of interest is defined as shown in formula (2):

$w_i=f\left(\frac{n_i\&times;\mathrm{\&alpha;}}{T-t_i+\mathrm{\&alpha;}}\right)$式(2)$w_i=f\left(\frac{{\mathrm{no}}_i\&times;\mathrm{\&alpha;}}{T-t_i+\mathrm{\&alpha;}}\right)$ Formula (2)

其中，n_i是该用户上网访问日志中c_i类日志的数量，t_i是该用户上网访问日志中c_i类日志的最近出现时间，T是当前系统时间，α＞0是一个调整系数。式(2)的意义是，由于用户的兴趣爱好会随时间的推移而动态变化，一些用户原本感兴趣的内容会随时间的推移而渐渐遗忘，新的兴趣会逐渐产生，因此用户的第i+1(0≤i＜m)类兴趣的权重(兴趣度)w_i与用户感兴趣的c_i类网页文档数成正比，与最近阅览c_i类网页文档的时间差成反比。该兴趣度相当于引入了用户兴趣动态更新机制，实现了对用户兴趣的稳定和准确更新，使得获得的用户兴趣爱好信息更加准确和及时，进一步提升用户体验，解决了用户兴趣漂移问题。Among them, n_i is the number of ci_- type logs in the user's online access log, t_i is the latest appearance time of the ci_- type log in the user's online access log, T is the current system time, and α>0 is an adjustment coefficient. The meaning of Equation (2) is that since the user’s interests and hobbies will change dynamically over time, some content that the user is originally interested in will gradually be forgotten over time, and new interests will gradually emerge, so the user’s i-th +1 (0≤i<m) The weight (degree of interest) w_i of category interest is directly proportional to the number of webpage documents of category c_i that the user is interested in, and inversely proportional to the time difference of recently browsing the webpage documents of category c_i . This degree of interest is equivalent to the introduction of a dynamic update mechanism for user interests, which realizes stable and accurate updates of user interests, makes the obtained user interest and hobbies information more accurate and timely, further improves user experience, and solves the problem of user interest drift.

其中，调整系统α值用于防止f(·)函数的分母为0，同时还可以调整兴趣度w_i的衰减速度，α值越小，兴趣度w_i的衰减速度越快，因此可以根据具体需求确定α的取值。Among them, adjusting the system α value is used to prevent the denominator of the f(·) function from being 0. At the same time, it can also adjust the decay speed of the interest degree w_i . The smaller the value of α, the faster the decay speed_of the interest degree wi. Need to determine the value of α.

为防止兴趣度数据过度偏向于上网积极用户(即日志数量较大用户)，因此需要正规化兴趣度数据。正规化后该用户的第i+1(0≤i＜m)类兴趣的权重(兴趣度)

定义如式(3)所示：In order to prevent the interest degree data from being excessively biased towards users who are actively surfing the Internet (that is, users with a large number of logs), it is necessary to normalize the interest degree data. The weight (degree of interest) of the user's i+1 (0≤i<m) type of interest after normalization

The definition is shown in formula (3):

$\stackrel{\&OverBar;}{w_i}=\frac{w_i}{\underset{k=0}{\overset{m-1}{\mathrm{\&Sigma;}}}{w}_k}$式(3)$\stackrel{\&OverBar;}{w_i}=\frac{w_i}{\underset{k=0}{\overset{m-1}{\mathrm{\&Sigma;}}}{w}_k}$ Formula (3)

其中，w_i由式(2)计算得到。Among them, w_i is calculated by formula (2).

用户兴趣学习模块会定期对用户兴趣模型进行更新，更新的方式即为根据上述的上网日志信息，如日志的主题、数量和时间等，对兴趣度进行更新。The user interest learning module will regularly update the user interest model by updating the interest degree according to the above-mentioned online log information, such as the subject, quantity and time of the log.

本实施例的电信业务推荐方法，通过根据用户的上网日志信息建立用户兴趣模型，以及用户兴趣和电信业务的关联关系等，可以据此生成推荐业务列表，对用户进行电信业务推荐，解决了电信业务推荐的成功率低的问题，实现了基于用户兴趣进行电信业务推荐，以及电信业务的精准营销，提升电信业务推荐的用户体验。The method for recommending telecommunications services in this embodiment establishes a user interest model based on the user's log information on the Internet, as well as the association between user interests and telecommunications services, etc., so as to generate a list of recommended services and recommend telecommunications services to users, solving the problem of telecommunications. The problem of the low success rate of service recommendation enables telecom service recommendation based on user interests and precise marketing of telecom services to improve the user experience of telecom service recommendation.

实施例六Embodiment six

图5为本发明电信业务推荐方法实施例中的获得用户兴趣和电信业务关联关系的流程示意图，如图5所示，本实施例可以包括以下步骤：FIG. 5 is a schematic flow diagram of obtaining user interests and telecommunication service associations in an embodiment of the method for recommending telecommunication services according to the present invention. As shown in FIG. 5 , this embodiment may include the following steps:

步骤501、电信业务推荐模块根据用户兴趣模型、电信业务信息和订购业务信息，建立二维带权记录集合；Step 501, the telecommunications service recommendation module establishes a two-dimensional weighted record set according to the user interest model, telecommunications service information and subscription service information;

例如，电信业务推荐系统根据用户的订购业务信息、电信业务信息以及用户兴趣模型库，建立二维带权记录集合T，其中，T＝{t₁，t₂，......，t_n}，其中n是系统内当前用户数；T中的每个记录t_u对应一个用户u的二维带权项集合

其中，m_u(0＜m_u≤|C|)是用户u的兴趣数量，|C|是系统主题标引类别总数，c_ui是用户u的第i+1(0≤i＜m_u)类兴趣名称，n_u(0＜n_u≤|S|)是该用户订购业务数量，|S|是电信业务总数，s_uj是用户u的第j+1(0≤j＜n_u)类电信业务名称，记录t_u对应的权重向量其中，w_ui是用户u正规化后c_ui类兴趣的权重(0≤i＜m_u)，p_uj是用户u订购业务s_uj的权重(0≤j＜n_u)，根据系统需求，这里令p_uj＝1。For example, the telecommunications service recommendation system establishes a two-dimensional weighted record set T according to the user's subscription service information, telecommunications service information and user interest model library, where T={t₁ , t₂ ,...,t_n }, where n is the current number of users in the system; each record t_u in T corresponds to a two-dimensional set of weighted items for a user u

Among them, m_u (0<m_u ≤|C|) is the number of interests of user u, |C| is the total number of system subject indexing categories, c_ui is the i+1th of user u (0≤i<m_u ) class interest name, n_u (0<n_u ≤|S|) is the number of services ordered by the user, |S| is the total number of telecommunications services, s_uj is the j+1 (0≤j<n_u ) class of user u Telecom business name, weight vector corresponding to record t_u Among them, w_ui is the weight of interest in c_ui after user u is normalized (0≤i<m_u ), p_uj is the weight of user u’s subscription service s_uj (0≤j<n_u ), according to system requirements, here Let p_uj =1.

步骤502、电信业务推荐模块计算所述集合中各类兴趣的支持度；Step 502, the telecommunications service recommendation module calculates the support degree of various interests in the set;

例如，电信业务推荐系统根据式(4)和式(5)计算二维带权记录集合T中各类兴趣C_k(0≤k＜|C|)的支持度Support(C_k)；For example, the telecom service recommendation system calculates the support degree Support(C_k ) of various interests C_k (0≤k<|C|) in the two-dimensional weighted record set T according to formula (4) and formula (5);

$\mathrm{Support}\left(C_k\right)=\frac{\underset{u}{\mathrm{\&Sigma;}}{\mathrm{Sup}}_{\mathrm{uk}}}{\left|T\right|}$式(4)$\mathrm{Support}\left(C_k\right)=\frac{\underset{u}{\mathrm{\&Sigma;}}{\mathrm{Sup}}_{\mathrm{uk}}}{\left|T\right|}$ Formula (4)

${\mathrm{Sup}}_{\mathrm{uk}}=\left\{\begin{array}{cc}{w}_{\mathrm{ui}},& \mathrm{if}{C}_k\&Element;t_u\mathrm{and}{C}_k=c_{\mathrm{ui}}\\ 0,& \mathrm{else}\end{array}\right.$式(5)${\mathrm{Sup}}_{\mathrm{uk}}=\left\{\begin{array}{cc}{w}_{\mathrm{ui}},& \mathrm{if}{C}_k\&Element;t_u\mathrm{and}{C}_k=c_{\mathrm{ui}}\\ 0,& \mathrm{else}\end{array}\right.$ Formula (5)

步骤503、电信业务推荐模块计算所述集合中的兴趣和业务二维项集的支持度；Step 503, the telecommunications service recommendation module calculates the support degree of the interest and service two-dimensional item sets in the set;

例如，电信业务推荐模块根据式(6)和式(7)计算二维带权记录集合T中{兴趣，业务}二维项集{C_k，S_l}(0≤k＜|C|，0≤l＜|S|)的支持度Support(C_k，S_l)：For example, the telecom service recommendation module calculates {interest, business} two-dimensional item set {C_k , S_l } (0≤k<|C|, 0≤l<|S|) Support (C_k , S_l ):

$\mathrm{Support}(C_k,S_l)=\frac{\underset{u}{\mathrm{\&Sigma;}}{\mathrm{Sup}}_{\mathrm{uq}}}{\left|T\right|}$式(6)$\mathrm{Support}(C_k,S_l)=\frac{\underset{u}{\mathrm{\&Sigma;}}{\mathrm{Sup}}_{\mathrm{uq}}}{\left|T\right|}$ Formula (6)

${\mathrm{Sup}}_{\mathrm{uq}}=\left\{\begin{array}{cc}{w}_{\mathrm{ui}},& \mathrm{if}\{C_k,S_l\}\&SubsetEqual;t_u\mathrm{and}{C}_k=c_{\mathrm{ui}}\\ 0,& \mathrm{else}\end{array}\right.$式(7)${\mathrm{Sup}}_{\mathrm{uq}}=\left\{\begin{array}{cc}{w}_{\mathrm{ui}},& \mathrm{if}\{C_k,S_l\}\&SubsetEqual;t_u\mathrm{and}{C}_k=c_{\mathrm{ui}}\\ 0,& \mathrm{else}\end{array}\right.$ Formula (7)

步骤504、电信业务推荐模块根据所述各类兴趣的支持度以及二维项集的支持度，得到用户兴趣和电信业务的关联规则的置信度。Step 504 , the telecommunication service recommendation module obtains the confidence level of the user interest and the association rule of the telecommunication service according to the support level of various interests and the support level of the two-dimensional itemset.

例如，电信业务推荐系统根据式(8)计算二维带权记录集合T中兴趣-业务关联规则

的置信度

并将其与规则

一起存储到兴趣-业务关联库；For example, the telecommunications service recommendation system calculates the interest-service association rule in the two-dimensional weighted record set T according to formula (8):

Confidence of

and combine it with the rule

Store together in the interest-business association library;

$\mathrm{Conf}(C_k\&DoubleRightArrow;S_l)=\left\{\begin{array}{cc}\frac{\mathrm{Support}(C_k,S_l)}{\mathrm{Support}\left(C_k\right)}& \mathrm{ifSupport}\left(C_k\right)\&NotEqual;0\\ 0,& \mathrm{else}\end{array}\right.$式(8)$\mathrm{conf}(C_k\&DoubleRightArrow;S_l)=\left\{\begin{array}{cc}\frac{\mathrm{Support}(C_k,S_l)}{\mathrm{Support}\left(C_k\right)}& \mathrm{ifSupport}\left(C_k\right)\&NotEqual;0\\ 0,& \mathrm{else}\end{array}\right.$ Formula (8)

本实施例的电信业务推荐方法，通过根据用户的上网日志信息建立用户兴趣模型，以及用户兴趣和电信业务的关联关系等，可以据此生成推荐业务列表，对用户进行电信业务推荐，解决了电信业务推荐的成功率低的问题，实现了基于用户兴趣进行电信业务推荐，以及电信业务的精准营销，提升电信业务推荐的用户体验。The method for recommending telecommunications services in this embodiment establishes a user interest model based on the user's log information on the Internet, as well as the association between user interests and telecommunications services, etc., so as to generate a list of recommended services and recommend telecommunications services to users, solving the problem of telecommunications. The problem of the low success rate of service recommendation enables telecom service recommendation based on user interests and precise marketing of telecom services to improve the user experience of telecom service recommendation.

实施例七Embodiment seven

图6为本发明电信业务推荐方法实施例中的电信业务推荐的流程示意图，如图6所示，本实施例可以包括以下步骤：FIG. 6 is a schematic flow diagram of the telecommunications service recommendation in the embodiment of the telecommunications service recommendation method of the present invention. As shown in FIG. 6, this embodiment may include the following steps:

步骤601，电信业务推荐模块根据用户兴趣模型库、兴趣-业务关联库和式(9)计算对用户u推荐其未订购业务S₁(0≤l＜|S|)的可信度

Step 601, the telecom service recommendation module calculates the credibility of recommending the unsubscribed service S₁ (0≤l<|S|) to user u according to the user interest model library, interest-service association library and formula (9)

$\mathrm{Confidence}(u\&DoubleRightArrow;S_l)=\underset{i=0}{\overset{m_u-1}{\mathrm{\&Sigma;}}}{w}_{\mathrm{ui}}\&CenterDot;\mathrm{Conf}(c_{\mathrm{ui}}\&DoubleRightArrow;S_l)$式(9)$\mathrm{Confidence}(u\&DoubleRightArrow;S_l)=\underset{i=0}{\overset{m_u-1}{\mathrm{\&Sigma;}}}{w}_{\mathrm{ui}}\&Center\; Dot;\mathrm{conf}(c_{\mathrm{ui}}\&DoubleRightArrow;S_l)$ Formula (9)

步骤602，电信业务推荐模块根据

将所有向用户u推荐未订购业务S₁进行降序排列，生成针对用户u的推荐业务列表，推送给用户。Step 602, the telecommunications service recommendation module according to

Arrange all recommended unsubscribed services_S1 to user u in descending order, generate a list of recommended services for user u, and push them to the user.

本实施例的电信业务推荐方法，通过根据用户的上网日志信息建立用户兴趣模型，以及用户兴趣和电信业务的关联关系等，可以据此生成推荐业务列表，对用户进行电信业务推荐，解决了电信业务推荐的成功率低的问题，实现了基于用户兴趣进行电信业务推荐，以及电信业务的精准营销，提升电信业务推荐的用户体验。The method for recommending telecommunications services in this embodiment establishes a user interest model based on the user's log information on the Internet, as well as the association between user interests and telecommunications services, etc., so as to generate a list of recommended services and recommend telecommunications services to users, solving the problem of telecommunications. The problem of the low success rate of service recommendation enables telecom service recommendation based on user interests and precise marketing of telecom services to improve the user experience of telecom service recommendation.

实施例八Embodiment Eight

在上述实施例的基础上，进一步的，对电信业务推荐的效果进行评估，并根据所述评估结果，优化所述电信业务推荐。On the basis of the foregoing embodiments, further, the effect of the telecommunications service recommendation is evaluated, and the telecommunications service recommendation is optimized according to the evaluation result.

通过采取推荐效果评估反馈和更新机制，实现流程的闭环工作，提高了系统的自适应能力。By adopting the recommendation effect evaluation feedback and update mechanism, the closed-loop work of the process is realized, and the adaptive ability of the system is improved.

例如，该评估反馈可以是以一段时间(周、月或季度等)为周期，观察向用户推荐其未订购业务S₁后，用户是否订购了该业务、业务推荐时间与订购业务开始时间的间隔等等，从而评价这些推荐的有效性和时效性，并可结合电信业务部门的营销策略，相应调整

For example, the evaluation feedback can be based on a period of time (week, month or quarter, etc.) to observe whether the user has ordered the service, the interval between the service recommendation time and the start time of the ordered service after recommending its unsubscribed service_S1 to the user. etc., so as to evaluate the effectiveness and timeliness of these recommendations, and adjust accordingly in combination with the marketing strategy of the telecom business department

例如，推荐评估模块根据式(10)和式(11)计算兴趣-业务关联规则

的适应度

For example, the recommendation evaluation module calculates interest-business association rules according to formula (10) and formula (11)

fitness

$\mathrm{Fit}(C_k\&DoubleRightArrow;S_l)=\frac{\underset{i=0}{\overset{N_u}{\mathrm{\&Sigma;}}}{H}_i}{N_u}$式(10)$\mathrm{fit}(C_k\&DoubleRightArrow;S_l)=\frac{\underset{i=0}{\overset{N_u}{\mathrm{\&Sigma;}}}{h}_i}{N_u}$ Formula (10)

式(11)

Formula (11)

其中，N_u是所有被推荐了业务S₁并且具有兴趣C_k的用户总数。TO_i是被推荐了业务S₁并且具有兴趣C_k的用户i开始订购业务S₁的时间，TR_i是向用户i推荐业务S₁的时间。显然有

Among them,_Nu is the total number of users who have been recommended the service S₁ and have interest C_k . TO_i is the time when user i who has been recommended service_S1 and has interest C_k starts to order service_S1 , and TR_i is the time when service_S1 is recommended to user i. apparently

推荐优化模块根据式(12)调整兴趣-业务关联规则的置信度The recommendation optimization module adjusts the interest-business association rules according to formula (12) Confidence of

$\mathrm{Conf}(C_k\&DoubleRightArrow;S_l)=\mathrm{Conf}(C_k\&DoubleRightArrow;S_l)*\mathrm{Fit}(C_k\&DoubleRightArrow;S_l)$式(12)$\mathrm{conf}(C_k\&DoubleRightArrow;S_l)=\mathrm{conf}(C_k\&DoubleRightArrow;S_l)*\mathrm{fit}(C_k\&DoubleRightArrow;S_l)$ Formula (12)

本领域普通技术人员可以理解：实现上述方法实施例的全部或部分步骤可以通过程序指令相关的硬件来完成，前述的程序可以存储于一计算机可读取存储介质中，该程序在执行时，执行包括上述方法实施例的步骤；而前述的存储介质包括：ROM、RAM、磁碟或者光盘等各种可以存储程序代码的介质。Those of ordinary skill in the art can understand that all or part of the steps for realizing the above-mentioned method embodiments can be completed by hardware related to program instructions, and the aforementioned program can be stored in a computer-readable storage medium. When the program is executed, the It includes the steps of the above method embodiments; and the aforementioned storage medium includes: ROM, RAM, magnetic disk or optical disk and other various media that can store program codes.

最后应说明的是：以上实施例仅用以说明本发明的技术方案，而非对其限制；尽管参照前述实施例对本发明进行了详细的说明，本领域的普通技术人员应当理解：其依然可以对前述各实施例所记载的技术方案进行修改，或者对其中部分技术特征进行等同替换；而这些修改或者替换，并不使相应技术方案的本质脱离本发明各实施例技术方案的精神和范围。Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: it can still be Modifications are made to the technical solutions described in the foregoing embodiments, or equivalent replacements are made to some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims (12)

1. a telecommunication service recommend method is characterized in that, comprising:

Obtain telecommunication service information, user's identity information, order business information and internet log information;

Obtain the subject indexing of user capture web document according to described internet log information; Subject indexing and internet log information according to described user capture web document are set up user interest model;

According to described user interest model, telecommunication service information and order business information, obtain the incidence relation of user interest and telecommunication service; And, obtain recommending service lists according to the incidence relation and the described user interest model of described user interest and telecommunication service, the user is carried out telecommunication service recommend.

2. telecommunication service recommend method according to claim 1 is characterized in that, obtains user capture web document subject indexing according to described internet log information, comprising:

Obtain the URL address that the user visits according to described internet log information, and, judge whether the user capture web document of this URL address correspondence has carried out subject indexing according to described URL address lookup URL-theme mapping library; If then directly obtain the subject indexing of described user capture web document;

Otherwise use webpage collector obtains the user capture web document of described URL correspondence, and described user capture web document is carried out preliminary treatment, obtains the text that standardizes; Described standardization text is carried out Chinese word segmentation and the identification of English phrase, remove the non-categorical information speech, obtain being with participle sign text; Described band participle sign text is carried out word frequency statistics and feature weight calculating, generate text vector; And described text vector carried out text classification, obtain described user capture web document subject indexing; Store described user capture web document subject indexing into described URL-theme mapping library.

3. telecommunication service recommend method according to claim 1 is characterized in that, sets up user interest model according to described user capture web document subject indexing and internet log information, comprising:

Comprise a plurality of user interest items in the described user interest model, comprise category of interest, the interest-degree corresponding in the described user interest item with described category of interest; The quantity of described user interest equals the quantity of described user capture web document subject indexing;

Described interest-degree is directly proportional with the daily record quantity corresponding with described category of interest in the described internet log information, is inversely proportional to the nearest time of occurrence with the corresponding daily record quantity of described category of interest in the described internet log information.

4. telecommunication service recommend method according to claim 3 is characterized in that, also comprises:

Described interest-degree is carried out normalization process, obtain regular interest-degree.

5. telecommunication service recommend method according to claim 1 is characterized in that, according to described user interest model, telecommunication service information and order business information, obtains the incidence relation of user interest and telecommunication service, comprising:

According to described user interest model, telecommunication service information and order business information, set up two-dimentional cum rights set of records ends;

Calculate the support of all kinds of interest in the described set; Calculate an interest and a professional two dimension support that collects in the described set;

According to the support of described all kinds of interest and the support of a two dimension collection, obtain the confidence level of the correlation rule of user interest and telecommunication service.

6. telecommunication service recommend method according to claim 5 is characterized in that, incidence relation and described user interest model according to described user interest and telecommunication service obtain recommending service lists, comprising:

According to the confidence level of the correlation rule of described user interest model and described user interest and telecommunication service, obtain the confidence level that telecommunication service is recommended;

Confidence level according to described telecommunication service is recommended with institute's professional arrangement to be recommended, generates described recommendation service lists.

7. according to the arbitrary described telecommunication service recommend method of claim 1-6, it is characterized in that, also comprise:

The effect that telecommunication service is recommended is assessed, and, optimized described telecommunication service and recommend according to described assessment result.

8. a telecommunication service commending system is characterized in that, comprising: user data acquisition module, user interest study module and telecommunication service recommending module;

Described user data acquisition module is used to obtain telecommunication service information, user's identity information, order business information and internet log information;

Described user interest study module is used for obtaining according to described internet log information the subject indexing of user capture web document; Subject indexing and internet log information according to described user capture web document are set up user interest model;

Described telecommunication service recommending module is used for according to described user interest model, telecommunication service information and order business information, obtains the incidence relation of user interest and telecommunication service; And, obtain recommending service lists according to the incidence relation and the described user interest model of described user interest and telecommunication service, the user is carried out telecommunication service recommend.

9. telecommunication service commending system according to claim 8 is characterized in that, described user interest study module comprises:

Judging unit is used for obtaining the URL address that the user visits according to described internet log information, and according to described URL address lookup URL-theme mapping library, judges whether the user capture web document of this URL address correspondence has carried out subject indexing;

First processing unit is used in the judged result of described judging unit directly obtaining described subject indexing when being;

Second processing unit is used in the judged result of described judging unit for not the time, uses the webpage collector to obtain the user capture web document of described URL correspondence, and described user capture web document is carried out preliminary treatment, and text obtains standardizing; Described standardization text is carried out Chinese word segmentation and the identification of English phrase, remove the non-categorical information speech, obtain being with participle sign text; Described band participle sign text is carried out word frequency statistics and feature weight calculating, generate text vector; And described text vector carried out text classification, obtain described user capture web document subject indexing; Store described user capture web document subject indexing into described URL-theme mapping library.

10. telecommunication service commending system according to claim 8 is characterized in that, described user interest study module also comprises:

The first interest unit, be used for setting up user interest model according to described user capture web document subject indexing and internet log information, comprise a plurality of user interest items in the described user interest model, comprise category of interest, the interest-degree corresponding in the described user interest item with described category of interest; The quantity of described user interest equals the quantity of described user capture web document subject indexing; Described interest-degree is directly proportional with the daily record quantity corresponding with described category of interest in the described internet log information, is inversely proportional to the nearest time of occurrence with the corresponding daily record quantity of described category of interest in the described internet log information;

The second interest unit is used for described interest-degree is carried out normalization process, obtains regular interest-degree.

11. telecommunication service commending system according to claim 8 is characterized in that, described telecommunication service recommending module comprises:

Aggregation units is used for setting up two-dimentional cum rights set of records ends according to described user interest model, telecommunication service information and order business information;

Processing unit is used for calculating the support of all kinds of interest of described set; Calculate an interest and a professional two dimension support that collects in the described set;

Associative cell is used for obtaining the confidence level of the correlation rule of user interest and telecommunication service according to the support of described all kinds of interest and the support of a two dimension collection;

Credible unit is used for the confidence level according to the correlation rule of described user interest model and described user interest and telecommunication service, obtains the confidence level that telecommunication service is recommended;

Recommendation unit is used for the confidence level according to described telecommunication service recommendation, with institute's professional arrangement to be recommended, generates described recommendation service lists.

12. arbitrary according to Claim 8-11 described telecommunication service commending system is characterized in that, also comprises:

Recommend evaluation module, be used for that the user is carried out the effect that telecommunication service recommends and assess, obtain assessment result;

Recommend optimal module, be used for, optimize described telecommunication service and recommend according to described assessment result.

Images