Summary of the invention
In view of this, the purpose of the embodiment of the present invention is to propose a kind of user's evaluation methodology for application and device,The evaluation of user can be realized for application, and then filter out the high-quality user of data balancing.
A kind of user's evaluation methodology for application provided based on the above-mentioned purpose embodiment of the present invention, including:
The application program that the user property label of statistical sample user and described sample of users are installed, calculates described applicationThe attribute trend data of program；
Obtain the service condition of each application program that targeted customer is installed, according to the service condition of described application programWith the attribute trend data of described application program, obtain the attribute evaluation of each application program that described targeted customer is installedValue；
According to the attribute evaluation value of each application program that described targeted customer is installed, obtain described targeted customer and pacifiedThe overall evaluation of the application program of dress, it is judged that whether described overall evaluation is more than presetting assessment threshold value, according to judged result, to instituteState targeted customer to be evaluated.
Optionally, described judge that whether described overall evaluation also includes more than before the step presetting assessment threshold value:
According to attribute classification known to described targeted customer, search the corresponding relation of attribute classification and the assessment threshold value presetList, obtains the default assessment threshold value that described targeted customer known attribute classification is corresponding.
Optionally, described judge that whether described overall evaluation also includes more than after the step presetting assessment threshold value:
If described overall evaluation is more than described assessment threshold value, then targeted customer is high-quality user；
If described overall evaluation is less than or equal to described assessment threshold value, then targeted customer is non-prime user；
Filter out high-quality user, and user property label corresponding for described high-quality user and described high-quality user are pacifiedThe application program of dress is as training data, and in the algorithm model of the user property prediction that input builds in advance, training is predictedThe forecast model of user property.
Optionally, the step of the described service condition obtaining each application program that targeted customer is installed also includes:
Obtaining the log information of mobile terminal reporting corresponding to targeted customer, statistics obtains each application program to be made every timeTime span；
Judge whether the time span that described application program is used every time is more than preset time threshold, be the most then designated asThe most effectively use, otherwise, be designated as the most invalid use；
Statistics obtains effective access times of each described application program, and by effective access times of each application programAccess times as each application program.
Optionally, the step of effective access times that described statistics obtains each application program also includes:
Statistics obtains effective access times of application program in the measurement period preset, it is judged that described application program effectiveWhether access times more than the frequency threshold value preset, the most then answer described default frequency threshold value as described in this cycleWith the access times of program, otherwise, the access times that effective access times are described application program of described application program；
Calculate each application program access times in the measurement period of predetermined number, and as each application programAccess times.
The embodiment of the present invention additionally provides a kind of user's evaluating apparatus for application, including:
Acquisition module, the application journey that the user property label and described sample of users for obtaining sample of users is installedSequence；And be sent to calculate mould by the application program that user property label and the described sample of users of described sample of users are installedBlock；
Computing module, for receiving the user property label of described sample of users that described acquisition module sends and describedThe application program that sample of users is installed, user property label and the described sample of users of adding up each described sample of users are pacifiedThe application program of dress, calculates the attribute trend data of described application program；The attribute trend data of described application program is sentTo attribute evaluation module；
Attribute evaluation module, for receiving the attribute trend data of the application program that described computing module sends, obtains meshThe service condition of each application program that mark user is installed, according to service condition and the described application program of described application programAttribute trend data, obtain the attribute evaluation value of each application program that described targeted customer is installed；By described each shouldIt is sent to evaluation module by the attribute evaluation value of program；
Evaluation module, for receiving the attribute evaluation value of each application program that described attribute evaluation module sends, according toThe attribute evaluation value of each application program that described targeted customer is installed, obtains the total of the application program that described user installedAssessed value, it is judged that whether described overall evaluation, more than presetting assessment threshold value, according to judged result, is commented described targeted customerValency.
Optionally, described evaluation module includes: user's evaluation module, threshold value search module and result judge module；
Described attribute evaluation module is additionally operable to, and the attribute evaluation value of described each application program is sent to user and assesses mouldBlock；
Described user's evaluation module, for receiving the attribute evaluation of each application program that described attribute evaluation module sendsValue, is calculated the application program overall evaluation based on known attribute classification that described targeted customer is installed；By described targetThe overall evaluation of user is sent to result judge module；
Described threshold value searches module, for according to attribute classification known to described targeted customer, searching the Attribute class presetThe other corresponding relation list with assessment threshold value, obtains the default assessment threshold value that described targeted customer known attribute classification is corresponding；WillDescribed default assessment threshold value is sent to result judge module；
Described result judge module, for receiving the overall evaluation of the targeted customer that described user's evaluation module sends, withAnd described threshold value searches the default assessment threshold value that module sends；Judge described targeted customer overall evaluation based on known attribute classificationWhether value is more than presetting assessment threshold value, the most then targeted customer is high-quality user；Otherwise, targeted customer is non-prime user.
Optionally, described user's evaluating apparatus also includes model training module；
Described model training module is used for, and filters out high-quality user, and by user property mark corresponding for described high-quality userThe application program that label and described high-quality user are installed is as training data, the calculation of the user property prediction that input builds in advanceIn method model, training obtains predicting the forecast model of user property.
Optionally, described attribute evaluation module also includes: Information Statistics module, time judgment module, number of times statistical moduleWith assessment computing module；
Described Information Statistics module, for obtaining the log information of mobile terminal reporting corresponding to targeted customer, adds upThe time span every time used to each application program；The time that each application program statistics obtained is used every time is longDegree is sent to time judgment module；
Described time judgment module, each application program sent for receiving described Information Statistics module is used every timeTime span, it is judged that whether the time span that application program is used every time more than preset time threshold, is the most then designated as oneSecondary effective use, otherwise, is designated as the most invalid use；
Described number of times statistical module, obtains effective access times of each application program for statistics, and by each applicationEffective access times of program are as the access times of each application program；The access times of each application program are sent to meterCalculate module；
Described assessment computing module, for receiving the use time of each application program that described number of times statistical module sendsNumber, and receive the attribute trend data of each application program that described computing module sends, by the use of each application programNumber of times is multiplied with the attribute trend data of described application program, is calculated each application program that described targeted customer is installedThe assessed value of corresponding different attribute classification；It is sent to the assessed value of described each application program correspondence different attribute classification evaluateModule.
Optionally, described number of times statistical module is additionally operable to, and statistics obtains the effective of application program in the measurement period presetAccess times, it is judged that whether effective access times of described application program are more than the frequency threshold value preset, the most then by described pre-If frequency threshold value as the access times of application program described in this cycle, otherwise, effectively using time of described application programNumber is the access times of described application program；It is calculated each application program use in the measurement period of predetermined number timeNumber, and as the access times of each application program.
From the above it can be seen that the user's evaluation methodology for application of embodiment of the present invention offer and device, logicalCross and be calculated the attribute trend data of each application program and the service condition of application program that targeted customer is installed, rootAccording to attribute trend data and the service condition of corresponding application program of each application program, it is calculated each application program correspondingThe assessed value of different attribute classification, then by based on known attribute classification for each application program of described targeted customer assessed valueIt is added, described targeted customer overall evaluation based on known attribute classification can be calculated, by by the general comment of targeted customerValuation assesses threshold ratio relatively with presetting, and finally can interpolate that whether targeted customer belongs to high-quality user.The described use for applicationFamily evaluation methodology and device can realize the evaluation of user for application, and then filter out the high-quality user of data balancing, it is possible toGreatly reduce the lack of uniformity of user data.
It should be appreciated that it is only exemplary and explanatory, not that above general description and details hereinafter describeThe present invention can be limited.
Detailed description of the invention
For making the object, technical solutions and advantages of the present invention clearer, below in conjunction with specific embodiment, and referenceAccompanying drawing, the present invention is described in more detail.
It should be noted that the statement of all uses " first " and " second " is for distinguishing two in the embodiment of the present inventionThe entity of individual same names non-equal or the parameter of non-equal, it is seen that " first " " second ", only for the convenience of statement, should notBeing interpreted as the restriction to the embodiment of the present invention, this is illustrated by subsequent embodiment the most one by one.
As a rule, in terminal, all of application program can regard have user property tendency as, such as, haveThe gender tendency of men and women, even if the tendency ratio adding up the data obtained last is 1:1, then can also regard its result as and haveGender tendency, simply the end value of tendency is impartial.But, between the different classes of middle different attribute classification of user propertyThere is the age attribute of huge difference, such as user in user data, uses the user of certain class application program to belong to 20-30 yearColony would generally be far longer than belong to 50-60 year colony, if using same standard to go selected user, then selectApplication program corresponding to sample of users will have bigger lack of uniformity so that last prediction, analysis result are inaccurate.PinTo this problem, embodiments provide a kind of user's evaluation methodology for application, with reference to shown in Fig. 1, for the present inventionThe schematic flow sheet of one embodiment of the user's evaluation methodology for application provided.The described user side of evaluation for applicationMethod, including:
Step 101, the application program that the user property label of statistical sample user and described sample of users are installed, calculateThe attribute trend data of described application program；
In order to evaluate the reliability of user, it is necessary first to acquire the characteristic number of the user data association corresponding with userAccording to, and present invention character of use based on mobile phone users and feature, find the most most frequently used and toolRepresentational event is exactly the use about types of applications program, and therefore, the present invention will be installed in the mobile terminal of userThe type of application program and the user property of user as evaluate user critical data.Described user property label refers toRepresent the label of user property or indicate the specific object classification of user property, such as: user property is sex, then describedUser property label is man or female；If user property is the age, the most described user property label is the concrete age or oneDetermine the age bracket (less than 20 years old, 20 years old to 30 years old, 30 years old to 40 years old, more than 40 years old etc.) of regular partition.Concrete, described useFamily attribute may include that sex, age, wedding be no, nationality, Income situation, geographical position, schooling (academic), religions beliefEtc. essential information.What the described here application program installed both may refer to that user installs on a mobile terminal shouldBy program, it is also possible to be the application program installed on multiple mobile terminals based on same user account of user.Described shiftingDynamic terminal may include that all kinds of Intelligent mobile equipment such as mobile phone, flat board.Described in this step obtain user property label and shouldBy program, it is common that pointer, for a large amount of different sample of users of known users attribute, obtains described a large amount of different samples respectivelyThe application program that user's correspondence is installed.
Optionally, in other embodiments of the present invention, described application program can also is that previously selected have userThe Application Type of attribute tendency.Such as: described user property is sex, then filter out in advance and there is answering of gender tendencyThe data used as needs by program, and other application programs without gender tendency (include the sample as training dataThe application program of this user and the application program of targeted customer) can ignore, so, it is possible not only to improve further predictionAccuracy, and speed and the efficiency of described user property prediction can be improved.
The application program that the sample of users of the known users attribute by obtaining is installed, it is possible to based on each application programStatistics obtains the trend data of different attribute classification.Wherein, described attribute trend data refers to be directed to a certain user propertyIn different attribute classification, the ratio data occupied respectively in each attribute classification.Such as: statistics obtains the age of known usersSection (less than 20 years old, 20 years old-29 years old, 30 years old-40 years old, more than 40 years old), and the application program installed of each sample of usersType, by above-mentioned data, can be calculated each application program number of users based on each age bracket of sample of users,And then it is calculated each application program tendency based on different age group ratio data, such as: for wechat user, 20The tendency ratio of the user that year is following is 8%, the tendency ratio of the user of 20 years old-29 years old is 35%, the user of 30 years old-40 years oldTendency ratio is 43%, the tendency ratio of the user of more than 40 years old is 14%.
Step 102, obtains the service condition of each application program that targeted customer is installed, according to described application programService condition and the attribute trend data of described application program, obtain the genus of each application program that described targeted customer is installedProperty assessed value；
According to the handling characteristics of mobile phone users, user uses the service condition of application program to be to evaluate user dependabilitySignificant data, by data acquisition, be calculated the use feelings of each application program that targeted customer to be evaluated is installedCondition.Described service condition typically refers to the access times of application program, uses frequency, use time etc. data.Such as: forFor each application program, the access times of each application program are multiplied can be calculated with corresponding attribute trend dataOne compound data, using described complex data as to each application program correspondence different attribute classification in the embodiment of the present inventionAssessed value, the access times of the application program that also will affect user data turn with attribute trend data the two important parameterTurn to concrete attribute evaluation value, and then be capable of the accurate evaluation to targeted customer.Here described attribute evaluation value is led toIt is often that sample of users is for without the other assessed value of Attribute class or scoring score value.
Step 103, according to the attribute evaluation value of each application program that described targeted customer is installed, obtains described targetThe overall evaluation of the application program that user is installed, it is judged that whether described overall evaluation is more than presetting assessment threshold value, according to judgementAs a result, described targeted customer is evaluated；
By the assessed value of each application program correspondence different attribute classification calculated in step 102, based on currentlyThe known attribute classification of targeted customer to be evaluated, by all application programs of current targeted customer to be evaluated based on this known attributeThe attribute evaluation value of classification is added, and can be calculated current targeted customer to be evaluated overall evaluation based on known attribute classificationValue.
Wherein, described default assessment threshold value refer to for evaluate that user data corresponding to user judges the most reliably oneIndividual demarcation line, both can be a calculated parameter threshold, it is also possible to be experiment, empirical numerical value.By settingAssessment threshold value, it is possible to insecure user is made a distinction with secure user, so make it possible to filter out have data balancing spyThe high-quality user of property.Described high-quality user refer to application program that described high-quality user uses with in the user property of this userThe other relatedness of Attribute class is more reliable, stable, it is possible to as training data based on APE user property, describedThe result that high-quality user makes the follow-up use for targeted customer, processes and evaluates is the most accurate, reliable.
From above-described embodiment, the user's evaluation methodology for application described in the embodiment of the present invention, by calculatingTo the service condition of the application program that attribute trend data and the current targeted customer to be evaluated of each application program are installed,The service condition of the attribute trend data according to each application program and corresponding application program, it is possible to be calculated each application journeyOrdered pair answers the assessed value of different attribute classification, then by each application program of current targeted customer to be evaluated based on known attributeThe assessed value of classification is added, and can be calculated current targeted customer to be evaluated overall evaluation based on known attribute classification, logicalCross and the overall evaluation of targeted customer is assessed threshold ratio relatively with presetting, finally can interpolate that whether described targeted customer belongs to high-qualityUser.Described user's evaluation methodology for application can realize the evaluation of user for application, and then filters out data balancingHigh-quality user, it is possible to greatly reduce the lack of uniformity of user data.
With reference to shown in Fig. 2, for the flow process of another embodiment for the user's evaluation methodology applied that the present invention providesSchematic diagram.Described user's evaluation methodology for application includes:
Step 201, the application program that the user property label of acquisition sample of users and described sample of users are installed；ItsIn, for evaluation objective user, needing first to determine the rule of evaluation, the embodiment of the present invention is applied by obtaining in sample of usersThe related data of program and after treatment, as the parameter of evaluation objective user.
Optionally, it is also possible to directly omit step 201, directly carry out step 202.
Step 202, adds up the user property label of each described sample of users and application that described sample of users is installedProgram, calculates the attribute trend data of each described application program；
Gathered user property label and the user of application program in this step and step 201, usually quantity is moreAnd the sample of users of known users attribute, the quantity of the sample of users of statistics is the most, and the trend data obtained is the most accurately and reliably.CanChoosing, what described attribute trend data referred to application program correspondence different attribute classification is specifically inclined to ratio data.
Step 203, obtains the log information of mobile terminal reporting corresponding to targeted customer, and statistics obtains each application programThe time span every time used；
In order to acquire the service condition of application program in mobile terminal, mobile terminal is needed user to be made in real timeOperation information or log information be reported in the server of background process, then background server is believed according to above-mentioned daily recordBreath statistics obtains the corresponding time span every time used of each application program that active user is installed.Above-mentioned report log informationProcess both can be the functional module carried in cell phone system or can also is that one of extra setting for gathering userThe application program of mobile phone of service condition or software.Above-mentioned current targeted customer to be evaluated both can be in step 201 and step 202Corresponding user, it is also possible to be other user.The user property of described current targeted customer to be evaluated is it is known that need to judge itData for application program and user property are the most reliable, namely whether described targeted customer is high-quality user.
Step 204, it is judged that whether the time span that application program is used every time is more than preset time threshold, the most thenPerform step 205, otherwise, perform step 206；
Use based on more application program has certain occasionality and promotes character, if using all of use timeNumber, then that can cause on access times is inaccurate, accordingly, it would be desirable to by the judgement with time threshold, filter out what essence usedThe access times of application program.Such as: certain number of site or application program, in order to promote or advertisement, are often boundDownload and install, the most often have the recommendation user installation of award and use this application program, and for some application programsUse for, its origin cause of formation used is not based on the custom of user oneself, but passive use, therefore can by this someThe access times of application program are got rid of.
Optionally, it is also possible to the use time interval setting application program filters out effective access times.Particularly as follows:When the use time interval of front and back twice is less than a certain default time interval threshold value, then can be by the use of twice before and after thisMerge into effective use once, be otherwise designated as twice effectively using.Such as: the first time of application program uses and makes with second timeTime interval be 3s, then be likely due to the disconnection of network or interruption that external force factor causes, the most stillBelong to use once.
Further, the access times of described application program can also combine with the time span used.Particularly as follows: statisticsThe use time span of the program that is applied, is multiplied with default weight coefficient and is calculated the time of this use of application programWeight, using time weighting as the coefficient of this use, finally can be calculated each application program with time weightingAccess times.In such manner, it is possible to the time span that user uses also serves as evaluating one of data of user, improve described forThe accuracy of user's evaluation methodology of application.
Step 205, according to step 204, the time span that described application program is used every time is more than preset time threshold,Represent that use time of described application program reaches the default standard time, so the use of described this time of application program being designated asThe most effectively use；
Step 206, according to step 204, the time span that described application program is used every time is less than or equal to when presettingBetween threshold value, represent the standard time that the use time of described application program is not reaching to preset, so should by described application programSecondary use is designated as the most invalid use；
Step 207, statistics obtains effective access times of application program in the measurement period preset, it is judged that described application journeyWhether effective access times of sequence more than the frequency threshold value preset, the most then perform step 208, otherwise, perform step 209；
Being directed to the use habit of different user, presumable user likes the intermittent application program that is used for multiple times, andSome users may like the most secondary long use, and the reliability of both use application program is the strongest, if simplyStatistics access times, can cause certain deviation.And, in the range of certain time, if the number of times used reaches certainThe upper limit, the numerical value of access times has had been out meaning, so needing to limit a frequency threshold value, limits the nothing of access timesLimit rises, and calculates to follow-up data and brings difference.When described measurement period refers to the cycle in user's use habit of restrictionBetween, it was generally a measurement period with one day.But described measurement period is not the whole time range of above-mentioned data statistics,Such as: with one day as measurement period, but the time range of statistics application program access times can be 1 month or 1 year etc.Deng.That is, in the range of the whole duration of statistics, multiple measurement period can be comprised.
Step 208, according to step 207, effective access times of described application program more than the frequency threshold value preset, then willDescribed default frequency threshold value is as the access times of application program described in this cycle；
Step 209, according to step 207, effective access times of described application program are less than or equal to the number of times threshold presetValue, the access times that effective access times are described application program of the most described application program；
Step 210, statistics obtains effective access times of each application program, and by effective use of each application programNumber of times is as the access times of each application program；
The access times of each application program are multiplied by step 211 respectively with the attribute trend data of this application program, meterCalculate the assessed value obtaining each application program correspondence different attribute classification that targeted customer the most to be evaluated is installed；
Step 212, according to attribute classification known to targeted customer, searches the attribute classification preset corresponding with assessment threshold valueRelation list, obtains the default assessment threshold value that described targeted customer known attribute classification is corresponding；
Described known attribute classification refers to that the specific object classification of user property is it is known that such as: the male in gender attributeOr women is known attribute classification.Although whether high-quality user can be belonged to according to assessment threshold decision targeted customer, butFor attribute classifications different in user property same for targeted customer, the difference between these attribute classifications is often large,If using same standard, then can cause the inaccurate of evaluation result.It is therefore desirable in advance for the spy of different attribute classificationLevy and set different assessment threshold values respectively, and then reach more optimal evaluation result, improve the accuracy that user evaluates.
Step 213, it is judged that whether current user to be evaluated overall evaluation based on known attribute classification is more than presetting assessmentThreshold value, the most then perform step 214, otherwise, perform step 215；
Step 214, according to step 213, current user to be evaluated overall evaluation based on known attribute classification is more than presettingAssessment threshold value, represents that the assessed value of user reaches default standard, therefore, it is determined that targeted customer the most to be evaluated uses for high-qualityFamily；
Step 215, according to step 213, targeted customer the most to be evaluated overall evaluation based on known attribute classification is littleIn or equal to presetting assessment threshold value, represent that the assessed value of targeted customer is not reaching to the standard preset, therefore, it is determined that described targetUser is non-prime user；
Optionally, can also include after step 215: according to attribute classification known to targeted customer, search the genus presetProperty classification and weight corresponding relation list, obtain targeted customer's weight based on different attribute classification；
When UAD and the application data of described high-quality user train forecast model as training dataTime, the biggest based on the difference between different attribute classification, it is therefore desirable to set correspondence respectively for different attribute classificationsWeighted value so that the result of calculation of last training pattern is the most accurate, avoids the data of the high-quality user filtered out to exist simultaneouslyHuge deviation or lack of uniformity, can be improved the harmony of training data, and then improve pre-by the different weight setSurvey the accuracy of result.
Step 216, uses weight corresponding to user property label corresponding for high-quality user, attribute tags and described targetThe Application Type that family is installed is as training data, in the algorithm model of the user property prediction that input builds in advance, instructionGet the forecast model of prediction user property.
By the high-quality user filtered out, it is possible to training obtains relatively reliable forecast model, and then can be more accuratePrediction obtain the user property of position user.
From above-described embodiment, the user's evaluation methodology for application described in the embodiment of the present invention is applied by judgementWhether the time span that program uses every time is more than the time threshold preset and judges the use in a measurement period timeNumber, whether beyond the frequency threshold value preset, adds up the program access times accurately and effectively that are applied, by for different genusProperty classification arranges different assessment threshold values and weight greatly reduces the difference between different attribute classification, finally by judgementWhether the assessed value of current targeted customer to be evaluated is more than the assessment threshold value preset and then filters out data reliable high-quality user.Meanwhile, described high-quality user is corresponding user property and the application program installed are as training data, enabling obtainIt is used for the forecast model of the user property of predicted position user the most accurately.Described user's evaluation methodology for application is very bigImprove the effectiveness of user application related data, reduce the other difference of Attribute class simultaneously, finally improve user and commentThe reliability and stability of valency.
It should be noted that above-described embodiment is intended merely to state an exemplary enforcement of the mentality of designing of the present inventionExample, and the thinking of the present invention is not limited to the quantity of step and the order stated in above-described embodiment.Namely it is directed to someStep can be omitted, or the order between the step having can also change as required, or can also be by above-mentioned twoStep in individual embodiment is combined using, and forms new embodiment.
In an optional embodiment of the present invention, described for application user's evaluation methodology can include step 201,Step 101, step 102, step 103, step 212, step 213, step 214, step 215.Particularly as follows:
The application program that the user property label of acquisition sample of users and described sample of users are installed；
Add up the user property label of each described sample of users and application program that described sample of users is installed, calculateThe attribute trend data of each application program；
Obtain the service condition of each application program that targeted customer is installed, according to the service condition of described application programWith the attribute trend data of described application program, obtain the attribute evaluation of each application program that described targeted customer is installedValue；
According to the attribute evaluation value of each application program that described targeted customer is installed, obtain described targeted customer and pacifiedThe overall evaluation of the application program of dress；
According to attribute classification known to targeted customer, search the corresponding relation row of attribute classification and the assessment threshold value presetTable, obtains the default assessment threshold value that current targeted customer to be evaluated known attribute classification is corresponding；
Judge that whether described targeted customer overall evaluation based on known attribute classification is more than presetting assessment threshold value, if so,Then judge described targeted customer as high-quality user, otherwise, it is determined that described targeted customer is non-prime user.
So, by for different attribute classification arrange different assessment threshold values can be substantially reduced different attribute classification itBetween diversity, improve user evaluate accuracy.
In another optional embodiment of the present invention, described user's evaluation methodology for application can include step201, step 101, step 102, step 103, step 213, step 214, step 215 and step 216.Particularly as follows:
The application program that the user property label of acquisition sample of users and described sample of users are installed；
Add up the user property label of each described sample of users and application program that described sample of users is installed, calculateThe attribute trend data of each application program；
Obtain the service condition of each application program that targeted customer is installed, according to the service condition of described application programWith the attribute trend data of described application program, obtain the attribute evaluation of each application program that described targeted customer is installedValue；
According to the attribute evaluation value of each application program that described targeted customer is installed, obtain described targeted customer and pacifiedThe overall evaluation of the application program of dress；
Judge that whether described targeted customer overall evaluation based on known attribute classification is more than presetting assessment threshold value, if so,Then judge described targeted customer as high-quality user, otherwise, it is determined that described targeted customer is non-prime user；
Filter out high-quality user, and user property label corresponding for high-quality user and described high-quality user are installedApplication Type is as training data, and in the algorithm model of the user property prediction that input builds in advance, training is predictedThe forecast model of user property.
Optionally, include after filtering out the step of high-quality user described in: according to attribute classification known to user, search pre-If the corresponding relation list of attribute classification and weight, obtain user's weight based on different attribute classification；By user based on notTraining data is also served as with the weight of attribute classification.
So, by user property label corresponding to the high-quality user that brush is selected and the application program installed asTraining data, it is possible to increase the accuracy of forecast model, and then make the prediction of user property the most accurate, reliable.
Meanwhile, by arranging different weights for different attribute classifications, it is possible to reduce the difference between attribute classification,Improve the balance between training data, finally improve accuracy and the reliability of forecast model.
In some optional embodiments of the present invention, described for application user's evaluation methodology can include step 201,Step 202, step 203, step 204, step 205, step 206, step 210, step 211, step 213, step 214, step215.Particularly as follows:
The application program that the user property label of acquisition sample of users and described sample of users are installed；
Add up the user property label of each described sample of users and application program that described sample of users is installed, calculateThe attribute trend data of each described application program；
Obtaining the log information of mobile terminal reporting corresponding to targeted customer, statistics obtains each application program to be made every timeTime span；
Judge whether the time span that application program is used every time is more than preset time threshold, be the most then designated as onceEffectively use, otherwise, be designated as the most invalid use；
Statistics obtains effective access times of each application program, and using effective access times of each application program asThe access times of each application program；
Respectively by the attribute trend data of the access times of each application program and this application program (namely user property pinTendency ratio for different attribute classification) it is multiplied, it is calculated each application program correspondence difference that targeted customer is installedThe other assessed value of Attribute class.
So, it is judged that whether each use time span of application program is more than the time threshold preset, and then adds upTo the access times of effective application program, improve accuracy and reliability that user evaluates.
In other optional embodiments of the present invention, described user's evaluation methodology for application can include step201, step 202, step 203, step 204, step 205, step 206, step 207, step 208, step 209, step 210,Step 211, step 213, step 214, step 215.Particularly as follows:
The application program that the user property label of acquisition sample of users and described sample of users are installed；
Add up the user property label of each described sample of users and application program that described sample of users is installed, calculateThe attribute trend data of each described application program；
Obtaining the log information of mobile terminal reporting corresponding to targeted customer, statistics obtains each application program to be made every timeTime span；
Judge whether the time span that application program is used every time is more than preset time threshold, be the most then designated as onceEffectively use, otherwise, be designated as the most invalid use；
Statistics obtains effective access times of application program in the measurement period preset, it is judged that described application program effectiveWhether access times more than the frequency threshold value preset, the most then answer described default frequency threshold value as described in this cycleWith the access times of program, otherwise, the access times that effective access times are described application program of described application program；
It is calculated each application program access times in the measurement period of predetermined number, and as each application journeyThe access times of sequence；
Access times according to each application program and the attribute trend data of this application program, be calculated the most to be evaluatedThe assessed value of each application program correspondence different attribute classification that the targeted customer of valency is installed；
Judge that whether described targeted customer overall evaluation based on known attribute classification is more than presetting assessment threshold value, if so,Then judge current targeted customer to be evaluated as high-quality user, otherwise, it is determined that targeted customer the most to be evaluated is non-prime user.
So, by being limited to the frequency threshold value of the upper limit of access times in a measurement period, it is possible to improve furtherThe validity and reliability of application program access times, and then the result that user is evaluated is the most reliable.
With reference to shown in Fig. 3, the structure for an embodiment of user's evaluating apparatus of application provided for the present invention is shownIt is intended to.Described user's evaluating apparatus for application, including:
Acquisition module 301, what the user property label and described sample of users for obtaining sample of users was installed shouldUse program；And the application program that user property label and the described sample of users of described sample of users are installed is sent to meterCalculate module 302；
Computing module 302, for receive described acquisition module 301 send described sample of users user property label withAnd application program, add up the user property label of each described sample of users and application program that described sample of users is installed,It is calculated the attribute trend data of each application program；Attribute trend data by calculated described each application programIt is sent to attribute evaluation module 303；
Attribute evaluation module 303, for receiving the attribute trend data of the application program that described computing module 302 sends,Calculate the service condition of each application program that targeted customer the most to be evaluated is installed, according to the use of each application programSituation and the attribute trend data of described application program, be calculated each application that targeted customer the most to be evaluated is installedThe assessed value of program correspondence different attribute classification；The assessed value of described each application program correspondence different attribute classification is sent toEvaluation module 304；
Evaluation module 304, for receiving the application program correspondence different attribute classification that described attribute evaluation module 303 sendsAssessed value, calculate all application programs overall evaluation based on known attribute classification that targeted customer the most to be evaluated is installedValue, it is judged that whether described targeted customer overall evaluation based on known attribute classification is more than presetting assessment threshold value, according to judging knotReally, described targeted customer is evaluated.Optionally, if overall evaluation is more than presetting assessment threshold value, the most current user to be evaluatedFor high-quality user；Otherwise, current user to be evaluated is non-prime user.
From above-described embodiment, described user's evaluating apparatus for application is calculated by described computing module 302To the attribute trend data of each application program, it is calculated each application program by described attribute evaluation module 303 correspondingThe assessed value of different attribute classification, by described evaluation module 304 be calculated that current targeted customer to be evaluated installed shouldBy program overall evaluation based on known attribute classification, then by by described overall evaluation with the assessment threshold ratio preset relatively,Judge that targeted customer the most to be evaluated, whether as high-quality user, finally filters out the high-quality for application data equilibrium and usesFamily.Described user's evaluating apparatus for application can realize the evaluation of user for application, and then filters out data balancingHigh-quality user, reduces the lack of uniformity of user data.
Shown in Figure 4, for the structure of another embodiment for the user's evaluating apparatus applied that the present invention providesSchematic diagram.
In some optional embodiments, described evaluation module 304 includes: user's evaluation module 3041, threshold value search mouldBlock 3042 and result judge module 3043；
Described attribute evaluation module 303 is additionally operable to, by the assessed value of described each application program correspondence different attribute classificationIt is sent to user's evaluation module 3041；
Described user's evaluation module 3041, for receiving each application program pair that described attribute evaluation module 303 sendsAnswer the assessed value of different attribute classification, be calculated all application programs that targeted customer the most to be evaluated installed based onKnow the other overall evaluation of Attribute class；The overall evaluation of user is sent to result judge module 3043；
Described threshold value searches module 3042, for according to attribute classification known to targeted customer the most to be evaluated, searchingThe attribute classification preset and the corresponding relation list of assessment threshold value, obtain current targeted customer to be evaluated known attribute classification correspondingDefault assessment threshold value；Described default assessment threshold value is sent to result judge module 3043；
Described result judge module 3043, for receiving the overall evaluation of the user that described user's evaluation module 3041 sendsValue, and the default assessment threshold value that described threshold value lookup module 3042 sends；Judge that targeted customer is based on known attribute classificationWhether overall evaluation is more than presetting assessment threshold value, and the most described targeted customer is high-quality user；Otherwise, described targeted customer isNon-prime user.
So, by for different attribute classification arrange different assessment threshold values can be substantially reduced different attribute classification itBetween diversity, improve user evaluate accuracy.
In another optional embodiment of the present invention, described user's evaluating apparatus also includes model training module 305；
Described model training module 305 is used for, and filters out high-quality user, and by user property label corresponding for high-quality userAnd the Application Type that described high-quality user is installed is as training data, the user property prediction that input builds in advanceIn algorithm model, training obtains predicting the forecast model of user property.So, by use that the high-quality user selected by brush is correspondingFamily attribute tags and the application program installed are as training data, it is possible to increase the accuracy of forecast model, and then makeThe prediction of user property is the most accurate, reliable.
In further embodiment of the present invention, described model training module 305 also includes: Weight Acquisition module 3051,Data screening module 3052 and data training module 3053；
Described Weight Acquisition module 3051, for according to attribute classification known to targeted customer, searching the Attribute class presetNot and the corresponding relation list of weight, targeted customer's weight based on different attribute classification is obtained；By described targeted customer based onThe weight of different attribute classification is sent to data training module 3053；
Described data screening module 3052, is used for filtering out high-quality user, and by user property mark corresponding for high-quality userThe Application Type that label and described high-quality user are installed is sent to data training module 3053；
Described data training module 3053, for receiving the user of described Weight Acquisition module 3051 transmission based on not belonging to togetherProperty the weight of classification and the user property label that sends of described data screening module 3052 and application program, by high-quality userWeight that corresponding user property label, attribute tags are corresponding and the Application Type conduct that described high-quality user is installedTraining data, in the algorithm model of the user property prediction that input builds in advance, training obtains predicting the prediction mould of user propertyType.
So, by arranging different weights for different attribute classifications, it is possible to reduce the difference between attribute classification,Improve the balance between training data, finally improve accuracy and the reliability of forecast model.
In some optional embodiments of the present invention, described attribute evaluation module 303 also includes: Information Statistics module3031, time judgment module 3032, number of times statistical module 3033 and assessment computing module 3034；
Described Information Statistics module 3031, for obtaining the day of mobile terminal reporting corresponding to current targeted customer to be evaluatedWill information, statistics obtains the time span that each application program is used every time；Each application program statistics obtained is eachThe time span used is sent to time judgment module 3032；
Described time judgment module 3032, every for receiving each application program of described Information Statistics module 3031 transmissionSecondary used time span, it is judged that whether the time span that application program is used every time is more than preset time threshold, if so,Then it is designated as the most effectively using, otherwise, is designated as the most invalid use；
Described number of times statistical module 3033, obtains effective access times of each application program for statistics, and by eachEffective access times of application program are as the access times of each application program；The access times of each application program are sentGive assessment computing module 3034；
Described assessment computing module 3034, for receiving each application program of described number of times statistical module 3033 transmissionAccess times, receive the attribute trend data of each application program that described ratio computing module 302 sends, answer each respectivelyIt is multiplied with the attribute trend data of described application program with the access times of program, is calculated targeted customer the most to be evaluatedThe assessed value of each application program correspondence different attribute classification installed；By described each application program correspondence different attribute classOther assessed value is sent to described evaluation module 304.
So, by judging that whether each use time span of application program is more than the time threshold preset, Jin ErtongMeter obtains the access times of effective application program, improves accuracy and reliability that user evaluates.
In further embodiment of the present invention, described number of times statistical module 3033 is additionally operable to, and statistics obtains the system presetEffective access times of application program in the meter cycle, it is judged that effective access times of described application program are the most secondary more than presetNumber threshold value, the most then using described default frequency threshold value as the access times of described application program in this cycle, otherwise, instituteState the access times that effective access times are described application program of application program；It is calculated each application program in present countAccess times in the statistics of variables cycle, and as the access times of each application program.
So, by being limited to the frequency threshold value of the upper limit of access times in a measurement period, it is possible to improve furtherThe validity and reliability of application program access times, and then the result that user is evaluated is the most reliable.
In still another aspect of the invention, additionally provide a kind of device, an embodiment of described device, including:
One or more processors, optionally, the one or more processor is used for performing any of the above one or manyThe step defined in method described in individual embodiment；And
For storing the memorizer of operational order；
The one or more processor is configured to from described memorizer obtain operational order and perform:
The application program that the user property label of statistical sample user and described sample of users are installed, calculates described applicationThe attribute trend data of program；
Obtain the service condition of each application program that targeted customer is installed, according to the service condition of described application programWith the attribute trend data of described application program, obtain the attribute evaluation of each application program that described targeted customer is installedValue；
According to the attribute evaluation value of each application program that described targeted customer is installed, obtain described targeted customer and pacifiedThe overall evaluation of the application program of dress, it is judged that whether described overall evaluation is more than presetting assessment threshold value, according to judged result, to instituteState targeted customer to be evaluated.
Optionally, it is characterised in that described judge that described overall evaluation is whether more than the step presetting assessment threshold value beforeAlso include:
According to attribute classification known to described targeted customer, search the corresponding relation of attribute classification and the assessment threshold value presetList, obtains the default assessment threshold value that described targeted customer known attribute classification is corresponding.
Optionally, described judge that whether described overall evaluation also includes more than after the step presetting assessment threshold value:
If described overall evaluation is more than described assessment threshold value, then targeted customer is high-quality user；
If described overall evaluation is less than or equal to described assessment threshold value, then targeted customer is non-prime user；
Filter out high-quality user, and user property label corresponding for described high-quality user and described high-quality user are pacifiedThe application program of dress is as training data, and in the algorithm model of the user property prediction that input builds in advance, training is predictedThe forecast model of user property.
Optionally, the step of the described service condition obtaining each application program that targeted customer is installed also includes:
Obtaining the log information of mobile terminal reporting corresponding to targeted customer, statistics obtains each application program to be made every timeTime span；
Judge whether the time span that described application program is used every time is more than preset time threshold, be the most then designated asThe most effectively use, otherwise, be designated as the most invalid use；
Statistics obtains effective access times of each described application program, and by effective access times of each application programAccess times as each application program.
Optionally, the step of effective access times that described statistics obtains each application program also includes:
Statistics obtains effective access times of application program in the measurement period preset, it is judged that described application program effectiveWhether access times more than the frequency threshold value preset, the most then answer described default frequency threshold value as described in this cycleWith the access times of program, otherwise, the access times that effective access times are described application program of described application program；
Calculate each application program access times in the measurement period of predetermined number, and as each application programAccess times.
Additionally, typically, the device described in the disclosure can be various electric terminal equipment, and such as mobile phone, individual digital helpsReason (PDA), panel computer (PAD), panel computer (PAD), intelligent television etc., therefore the protection domain of the disclosure should not limit intoCertain certain types of device.
Additionally, be also implemented as the computer program performed by CPU, this computer program according to disclosed methodCan store in a computer-readable storage medium.When this computer program is performed by CPU, perform disclosed method limitsFixed above-mentioned functions.
Additionally, said method step and system unit can also utilize controller and make controller real for storageThe computer-readable recording medium of the computer program of existing above-mentioned steps or Elementary Function realizes.
In addition, it should be appreciated that computer-readable recording medium as herein described (such as, memorizer) can be volatileProperty memorizer or nonvolatile memory, or volatile memory and nonvolatile memory can be included.As exampleSon and nonrestrictive, nonvolatile memory can include read only memory (ROM), programming ROM (PROM), electrically programmableROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory.Volatile memory can include random access memoryMemorizer (RAM), this RAM can serve as external cache.Nonrestrictive as an example, RAM can be with manyThe form of kind obtains, such as synchronous random access memory (DRAM), dynamic ram (DRAM), synchronous dram (SDRAM), double data rate SDRAM(DDR SDRAM), enhancing SDRAM (ESDRAM), synchronization link DRAM (SLDRAM) and direct RambusRAM (DRRAM).InstituteThe storage device of disclosed aspect is intended to include but not limited to the memorizer of these and other suitable type.
Those skilled in the art will also understand is that, in conjunction with the various illustrative logical blocks described by disclosure herein, mouldBlock, circuit and algorithm steps may be implemented as electronic hardware, computer software or a combination of both.Hard in order to clearly demonstratePart and this interchangeability of software, it is entered by the function with regard to various exemplary components, square, module, circuit and stepGo general description.This function is implemented as software and is also implemented as hardware and depends on specifically applying and applyingTo the design constraint of whole system.Those skilled in the art can realize described for every kind of concrete application in every wayFunction, but this realization decision should not be interpreted as causing a departure from the scope of the present disclosure.
Can utilize in conjunction with various illustrative logical blocks, module and the circuit described by disclosure herein and be designed toThe following parts performing function described here realize or perform: general processor, digital signal processor (DSP), special collectionBecome circuit (ASIC), field programmable gate array (FPGA) or other PLD, discrete gate or transistor logic, divideVertical nextport hardware component NextPort or any combination of these parts.General processor can be microprocessor, but alternatively, processesDevice can be any conventional processors, controller, microcontroller or state machine.Processor can also be implemented as calculating equipmentCombination, such as, DSP and the combination of microprocessor, multi-microprocessor, one or more microprocessor combine DSP core or anyOther this configuration.
Step in conjunction with the method described by disclosure herein or algorithm can be directly contained in hardware, be held by processorIn the software module of row or in combination of the two.Software module may reside within RAM memory, flash memory, ROM storageDevice, eprom memory, eeprom memory, depositor, hard disk, removable dish, CD-ROM or known in the art any itsIn the storage medium of its form.Exemplary storage medium is coupled to processor so that processor can be from this storage mediumMiddle reading information or to this storage medium write information.In an alternative, described storage medium can be with processor collectionBecome together.Processor and storage medium may reside within ASIC.ASIC may reside within user terminal.A replacementIn scheme, processor and storage medium can be resident in the user terminal as discrete assembly.
In one or more exemplary design, described function can be real in hardware, software, firmware or its combination in anyExisting.If realized in software, then described function can be stored in computer-readable as one or more instructions or codeTransmit on medium or by computer-readable medium.Computer-readable medium includes computer-readable storage medium and communication media,This communication media includes any medium contributing to that computer program is sent to another position from a position.Storage mediumIt can be any usable medium that can be accessed by a general purpose or special purpose computer.Nonrestrictive as an example, this computerComputer-readable recording medium can include RAM, ROM, EEPROM, CD-ROM or other optical disc memory apparatus, disk storage equipment or other magneticProperty storage device, or may be used for carrying or storage form is instruction or the required program code of data structure and canOther medium any accessed by universal or special computer or universal or special processor.Additionally, any connection canIt is properly termed as computer-readable medium.Such as, if using coaxial cable, optical fiber cable, twisted-pair feeder, digital subscriber line(DSL) or the wireless technology of such as infrared ray, radio and microwave is come from website, server or other remote source send software,The most above-mentioned coaxial cable, optical fiber cable, twisted-pair feeder, DSL or the most infrared are first, the wireless technology of radio and microwave is included inThe definition of medium.As used herein, disk and CD include compact disk (CD), laser disk, CD, digital versatile disc(DVD), floppy disk, Blu-ray disc, wherein disk the most magnetically reproduces data, and CD reproduces data with utilizing laser optics.OnThe combination stating content should also be as being included in the range of computer-readable medium.
Disclosed exemplary embodiment, but disclosed exemplary embodiment should be noted, it should be noted that without departing substantially fromOn the premise of the scope of the present disclosure that claim limits, may be many modifications and revise.According to disclosure described hereinThe function of the claim to a method of embodiment, step and/or action are not required to perform with any particular order.Although additionally, these public affairsThe element opened can describe or requirement with individual form, it is also contemplated that multiple, it is odd number unless explicitly limited.
It should be appreciated that it is used in the present context, unless exceptional case, singulative " clearly supported in contextIndividual " it is intended to also include plural form.It is to be further understood that "and/or" used herein refers to one or oneArbitrarily and likely combining of the individual above project listed explicitly.
Above-mentioned disclosure embodiment sequence number, just to describing, does not represent the quality of embodiment.
One of ordinary skill in the art will appreciate that all or part of step realizing above-described embodiment can pass through hardwareCompleting, it is also possible to instruct relevant hardware by program and complete, described program can be stored in a kind of computer-readableIn storage medium, storage medium mentioned above can be read only memory, disk or CD etc..