CN109086158A

CN109086158A - A kind of Analysis on Abnormal method, apparatus and server

Info

Publication number: CN109086158A
Application number: CN201810994869.0A
Authority: CN
Inventors: 游九龙
Original assignee: Tencent Technology Shenzhen Co Ltd
Current assignee: Tencent Technology Shenzhen Co Ltd
Priority date: 2018-08-29
Filing date: 2018-08-29
Publication date: 2018-12-25
Anticipated expiration: 2038-08-29
Also published as: CN109086158B

Abstract

The application provides a kind of Analysis on Abnormal method, apparatus, server and storage medium, this method comprises: determining the target recommendation results of the presence exception in recommender system；It obtains the recommender system and generates the destination request responded when the target recommendation results；From the historical execution information for different requests, it inquires the recommender system and responds the destination request and generate the target execution information generated when the target recommendation results, the historical execution information requested for one is that current time foregoing description recommender system responds the request and generates the execution information for requesting to generate when corresponding recommendation results；It is analyzed based on the reason of the target execution information generates the target recommendation results to the recommender system, obtains the abnormal cause of the recommender system.The application analysis on the basis of true reappearance recommender system generates target recommendation results obtains the abnormal cause of recommender system, improves the accuracy of the recommender system abnormal cause analyzed.

Description

A kind of Analysis on Abnormal method, apparatus and server

Technical field

This application involves recommender system technical fields, more particularly to a kind of Analysis on Abnormal method, apparatus and service Device.

Background technique

Although the appearance of internet and popularizing and bringing a large amount of information to user, user is met in the information age to letter The demand of breath, but when usually there is user in face of bulk information, it quickly can not therefrom obtain oneself interested information Problem.In order to solve this problem, its interested letter can be recommended to user according to Characteristic of Interest, behavioural habits of user etc. The recommender system of breath is come into being.

In order to guarantee recommender system generate recommendation results accuracy, it usually needs recommender system generate recommendation knot When fruit has abnormal (for example, recommendation results are not the interested information of user), the original of the recommendation results is generated to recommender system Because being analyzed, to be improved to recommender system based on the reason of analyze, so that the recommendation results that recommender system generates are more It is accurate to add.

The prior art be usually determine recommender system generate recommendation results deposit when abnormal, construction diagnostic analysis request And analog access recommender system is requested according to diagnostic analysis, to analyze the reason of recommender system is abnormal.However, because construction The scene of diagnostic analysis request analog access recommender system is not the scene that true recommender system generates abnormal recommendation results, So the prior art is requested in a manner of analyzing the reason of recommender system is abnormal construction diagnostic analysis, often exists and divide The problem of the recommender system abnormal cause inaccuracy of precipitation.

Summary of the invention

In view of this, the application provides a kind of Analysis on Abnormal method, apparatus, server and storage medium, to improve The accuracy of recommender system Analysis on Abnormal.

To achieve the goals above, technical solution provided in an embodiment of the present invention is as follows:

A kind of Analysis on Abnormal method, comprising:

Determine the target recommendation results of the presence exception in recommender system；

It obtains the recommender system and generates the destination request responded when the target recommendation results；

From the historical execution information for different requests, inquires the recommender system and respond destination request generation institute The target execution information generated when stating target recommendation results, the historical execution information requested for one is institute before current time It states recommender system and responds the request generation execution information for requesting to generate when corresponding recommendation results；

It is analyzed based on the reason of the target execution information generates the target recommendation results to the recommender system, Obtain the abnormal cause of the recommender system.

Preferably, the target recommendation results are to receive user to determine the selection operation of the recommendation results in recommender system , the acquisition recommender system generates the destination request responded when the target recommendation results, comprising:

It is asked from target is inquired before current time at least one request that the recommender system is sent from the user It asks, it is the target recommendation results that the recommender system, which responds the recommendation results that the destination request generates,.

Preferably, it is described from before current time from the user at least one request that the recommender system is sent Inquire destination request, comprising:

Determine that the user identifier of the user and the recommender system generate the time ranges of the target recommendation results；

Inquiry is identified with the target user and is closed from each request that current time foregoing description recommender system receives The request of connection, the recommender system each of receive the request and are associated with the use that the oriented recommender system sends the request The user identifier at family；

Destination request is determined from the request inquired, the recommender system responds the recommendation knot that the destination request generates Fruit is the target recommendation results, and the recommender system responds the time that the destination request generates the target recommendation results In the time range.

Preferably, it in the historical execution information from for different requests, inquires the recommender system and responds the mesh Mark request generates the target execution information generated when the target recommendation results, comprising:

Determine the historical execution information for being directed to different requests；The historical execution information requested for one, the request The historical execution information of request mark and the request；

From the historical execution information for different requests, the associated request for requesting to be identified as the destination request is inquired The historical execution information of mark, and the historical execution information inquired is responded into the target as the recommender system and is asked The target execution information generated when seeking survival into the target recommendation results.

Preferably, the recommender system responds the process packet that execution information is generated when the request generates the recommendation results It includes:

When obtaining the recommender system and responding the request generation recommendation results, the bandwagon effect log and property of generation It can log；The bandwagon effect log is the log that the recommender system generates when showing the recommendation results, the performance day Will is the log for constituting each service creation of the recommender system；

The bandwagon effect log and the performance logs are handled respectively, the bandwagon effect log that obtains that treated And performance logs；

Based on treated bandwagon effect log and the performance logs, execution information is generated.

Preferably, described that the bandwagon effect log and the performance logs are handled respectively, obtain that treated Bandwagon effect log and performance logs, comprising:

The bandwagon effect log and performance logs are stored in message queue；

Starting first consumption program and second consumption program, by it is described first consumption program to the bandwagon effect log into Row processing, the bandwagon effect log that obtains that treated, and the performance logs are carried out by the second consumption program Processing, the performance logs that obtain that treated.

Preferably, the bandwagon effect log includes user behavior information and user's portrait, and the recommender system responds institute It states and generates the process of the bandwagon effect log when request generates the recommendation results and include:

When obtaining the recommender system and responding the request generation recommendation results, the user for sending the request is current User behavior information and user portrait；

It is drawn a portrait based on the user behavior information got and user and generates the bandwagon effect log.

Preferably, the user behavior information of user includes each of sending to ask respectively to the recommender system with the user Relevant user is asked to request behavioural information, this method further include:

It determines before being requested described in the recommender system and is requested by the nearest preset quantity that the user sends；

User relevant to identified each request respectively is filtered out from the user behavior information got to request to go For information；

Described drawn a portrait based on the user behavior information got and user generates the bandwagon effect log, comprising: is based on The user's portrait got and the user filtered out request behavioural information to generate the bandwagon effect log.

Preferably, each service of the recommender system includes Broker service, Merger service and the service of recalling, institute Performance logs are stated to include Broker serve log, Merger serve log and recall serve log, in which:

The serve log of recalling is that the service of recalling is based on pushing away described in the recommender system response request generation When recommending result, the user of the request current user behavior information and user's portrait are sent, first is recalled from resource pool and is pushed away When recommending result, the log for being used to characterize the treatment process for recalling service of generation；

The Merger serve log is that Merger service is handled first recommendation results to obtain the When two recommendation results, the log of the treatment process for characterizing the Merger service of generation；

The Broker serve log is that the Broker service handles to obtain second recommendation results When stating recommendation results, the log of the treatment process for characterizing the Broker service of generation.

A kind of Analysis on Abnormal device, comprising:

Target recommendation results determination unit, for determining the target recommendation results of the exception of the presence in recommender system；

Destination request acquiring unit, for obtaining the mesh responded when the recommender system generates the target recommendation results Mark request；

Execution information query unit, for inquiring the recommender system from the historical execution information for different requests It responds the destination request and generates the target execution information generated when the target recommendation results, the history requested for one is held Row information is that current time foregoing description recommender system generates when responding the request generation corresponding recommendation results of the request Execution information；

Analysis on Abnormal unit is pushed away for generating the target to the recommender system based on the target execution information The reason of recommending result is analyzed, and the abnormal cause of the recommender system is obtained.

Preferably, the target recommendation results are to receive user to determine the selection operation of the recommendation results in recommender system , the destination request acquiring unit is specifically used for: sending before current time from the user to the recommender system Destination request is inquired at least one request, it is the mesh that the recommender system, which responds the recommendation results that the destination request generates, Mark recommendation results.

Preferably, the destination request acquiring unit, comprising:

Time range determination unit, user identifier and the recommender system for determining the user generate the mesh Mark the time range of recommendation results；

Query unit is requested, for inquiry and institute from each request that current time foregoing description recommender system receives It states target user and identifies associated request, the oriented recommender system of the request association that each of the recommender system receives Send the user identifier of the user of the request；

Destination request obtains subelement, for determining destination request, the recommender system response from the request inquired The recommendation results that the destination request generates are the target recommendation results, and the recommender system responds the destination request life It is located in the time range at the time of the target recommendation results.

Preferably, the execution information query unit, comprising:

Historical execution information determination unit, for determining the historical execution information for being directed to different requests；It is requested for one Historical execution information, request mark and the historical execution information of the request of the request；

Execution information inquires subelement, for inquiring associated request from the historical execution information for different requests It is identified as the historical execution information of the request mark of the destination request, and using the historical execution information inquired as institute It states recommender system and responds the destination request and generate the target execution information generated when the target recommendation results.

A kind of server, comprising: at least one processor and at least one processor；The memory is stored with program, The processor calls the program of the memory storage, and described program is for realizing the Analysis on Abnormal method.

A kind of storage medium, for storing the program for realizing the Analysis on Abnormal method.

A kind of Analysis on Abnormal method, apparatus, server and storage medium provided by the invention are each asked by being directed to Storage recommender system is asked to respond the mode that the request generates the execution information generated when recommendation results, so that depositing in recommender system At target recommendation results (the target recommendation results are abnormal recommendation results), which can be generated based on recommender system and pushed away The destination request responded when recommending result uniquely determines out from multiple execution informations of storage and really generates the target in history and push away The execution information generated when recommending result；And then based on the reason of the execution information generates the target recommendation results to recommender system into Row analysis, obtains the abnormal cause of recommender system.

Because the present invention is the execution information generated when really generating target recommendation results based on recommender system, to recommender system The analysis that the reason of generating the target recommendation results carries out；Construction diagnostic analysis, which is based on, accordingly, with respect to the prior art asks modulus For quasi- access recommender system is in a manner of determining recommender system abnormal cause, the present invention can be raw in true reappearance recommender system On the basis of target recommendation results, the reason of generating the target recommendation results to recommender system, is analyzed, and obtains recommending system The abnormal cause of system, and then improve the accuracy of the recommender system abnormal cause analyzed.

Detailed description of the invention

In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below There is attached drawing needed in technical description to be briefly described, it should be apparent that, the accompanying drawings in the following description is only this The embodiment of invention for those of ordinary skill in the art without creative efforts, can also basis The attached drawing of offer obtains other attached drawings.

Fig. 1 is a kind of composed structure schematic diagram of recommender system provided by the embodiments of the present application；

Fig. 2 is the composed structure schematic diagram of another recommender system provided by the embodiments of the present application；

Fig. 3 is a kind of hardware block diagram of Analysis on Abnormal server provided by the embodiments of the present application；

Fig. 4 is a kind of application scenarios schematic diagram of Analysis on Abnormal method provided by the embodiments of the present application；

Fig. 5 is the application scenarios schematic diagram of another Analysis on Abnormal method provided by the embodiments of the present application；

Fig. 6 is a kind of composed structure schematic diagram of recommendation server provided by the embodiments of the present application；

Fig. 7 is a kind of generation method flow chart of execution information provided by the embodiments of the present application；

Fig. 8 is that a kind of pair of bandwagon effect log provided by the embodiments of the present application and performance logs are handled, and is handled The application scenarios schematic diagram of bandwagon effect log and performance logs afterwards；

Fig. 9 is that a kind of pair of bandwagon effect log provided by the embodiments of the present application and performance logs are handled, and is handled The method flow diagram of bandwagon effect log and performance logs afterwards；

Figure 10 is a kind of Analysis on Abnormal method flow diagram provided by the embodiments of the present application；

Figure 11 be it is provided by the embodiments of the present application it is a kind of before current time from user to recommender system send at least The method flow diagram of destination request is inquired in one request；

Figure 12 is the application scenarios schematic diagram of another Analysis on Abnormal method provided by the embodiments of the present application；

Figure 13 is a kind of structural block diagram of Analysis on Abnormal device provided by the embodiments of the present application.

Specific embodiment

Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete Site preparation description, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.It is based on Embodiment in the present invention, it is obtained by those of ordinary skill in the art without making creative efforts every other Embodiment shall fall within the protection scope of the present invention.

The Analysis on Abnormal method of the application, when can be applied to there are abnormal recommendation results in recommender system, The reason of generating abnormal recommendation results to recommender system is analyzed, and analysis result is former as the exception of recommender system Cause, in order to which the abnormal cause based on recommender system improves recommender system, so that improved recommender system is subsequent In application process, recommendation results generated are more accurate, are more able to satisfy the demand of user, and then improve user to the recommendation system The viscosity of system.Wherein, the abnormal recommendation results in recommender system can generate recommendation results for recommender system and to user's exhibition After showing, by its unsatisfied recommendation results of user feedback.

In order to make it easy to understand, first the composed structure of recommender system provided by the embodiments of the present application is described in detail.Such as It may include terminal 11 and recommendation server 12 in recommender system shown in Fig. 1 shown in Fig. 1.

Wherein, terminal 11 can be mobile phone, tablet computer etc..

In the embodiment of the present application, it can be equipped in terminal 11 and recommend application program (Application, APP), be based on The APP, terminal can be established with recommendation server 12 belonging to the APP and be communicated to connect, and be asked to the recommendation server 12 transmission It asks.

Correspondingly, the request that recommendation server 12 is sent in response to terminal 11, returns to the APP by the APP in terminal and asks The data asked, wherein the request can be the request that request recommendation server carries out information recommendation, and recommendation server is into terminal The requested data of the APP that the APP is returned can be recommendation results.

For example, terminal is by recommending application program to send to recommendation server for requesting recommendation server to recommend article Request, after recommendation server receives request, can respond the request and generate recommendation results (includes at least one in the recommendation results Piece article), and the APP returns to the recommendation results into terminal.

Further, the recommendation server in recommender system provided by the embodiments of the present application generates recommendation results in respond request During, possible application to the current user behavior information for the user for sending request and user's portrait.

Wherein, the user behavior information of user may include: whenever user sends a request to recommender system, for this After requesting recommender system to return to recommendation results to user, feedback information of the user to recommendation results.

User can be what user was actively entered to the feedback information of recommendation results, may not be user be actively entered and It is to be automatically generated according to access situation of the user to recommendation results, which may include that user looks into recommendation results Read situation, satisfaction, interest level etc..

It is above only the preferred content of user behavior information provided by the embodiments of the present application, relevant user behavioural information Particular content, inventor can be arbitrarily arranged according to their own needs, it is not limited here.

User's portrait be according to the behavior of some primary attributes (including age, gender etc.) of user and user to user into Some information of row description, can be some labels of the hobby of reflection user.The use of the user portrait and the user of user Family behavioural information is related, and the user behavior information of user is updated with access situation of the user to recommender system (for example, every When user sends a request to recommender system, the user behavior information of user just increases a feedback information, the feedback information It is feedback information of the user to the recommendation results after recommender system responds request generation recommendation results), user's portrait of user It is to be updated with the update of the user behavior information of user.

In the embodiment of the present application, it is preferred that the user to user that can be realized in the memory of recommendation server The storage of behavioural information and user's portrait rings in order to which recommendation server is after receiving the request that user is sent by terminal User behavior information and user's portrait that the current user is locally obtained from recommendation server should be requested, and based on the use got Family behavioural information and user, which draw a portrait, generates recommendation results, and feeds back the recommendation results to terminal.

It is referring to fig. 2 the structural schematic diagram of another recommender system provided by the embodiments of the present application.It is pushed away in as shown in Figure 2 A data storage device 13 independently of recommendation server can also be additionally arranged by recommending in system, to realize the user to user The storage of behavioural information and user's portrait.In this way, when recommendation server receive request that user sent by terminal (referring to Step S101) after, the user behavior information and user that can obtain the current user from data storage device with respond request are drawn As (referring to step 102), and being drawn a portrait based on the user behavior information got and user and generating recommendation results (referring to step S103), and to terminal feed back the recommendation results (referring to step S104).

In the embodiment of the present application, it is preferred that data storage device 13 not only has data storage function but also can be with Has the function of data query, for example, data storage device can be user behavior inquiry system etc..

A kind of Analysis on Abnormal method provided by the embodiments of the present application can be applied to server, which can be Network side provides the service equipment of service for user, may be the server cluster of multiple servers composition, it is also possible to single Platform server.

The server that the Analysis on Abnormal method is applied to can be identical as recommendation server (at this point, recommendation service Information recommendation function had both may be implemented in device, was also able to achieve the Analysis on Abnormal function to the recommender system belonging to itself)；It can also To be independently of another server other than recommendation server (at this point, recommendation server is for realizing information recommendation function, independence Another server other than recommendation server can be used to implement the abnormal cause point of recommender system belonging to recommendation server Analyse function).

Optionally, Fig. 3 shows the hardware block diagram for the server that Analysis on Abnormal method is applied to, referring to figure 3, the hardware configuration of the server may include: processor 31, communication interface 32, memory 33 and communication bus 34；

In embodiments of the present invention, processor 31, communication interface 32, memory 33, communication bus 34 quantity can be with For at least one, and processor 31, communication interface 32, memory 33 complete mutual communication by communication bus 34；

Processor 31 may be a central processor CPU or specific integrated circuit ASIC (Application Specific Integrated Circuit), or be arranged to implement the integrated electricity of one or more of the embodiment of the present invention Road etc.；

Memory 33 may include high speed RAM memory, it is also possible to further include nonvolatile memory (non-volatile Memory) etc., a for example, at least magnetic disk storage；

Wherein, memory is stored with program, the program that processor can call memory to store, and program is used for:

It obtains recommender system and generates the destination request responded when target recommendation results；

From the historical execution information for different requests, inquiry recommender system response destination request generates target and recommends knot The target execution information generated when fruit, recommender system response is asked before the historical execution information requested for one is current time Seek survival into the execution information for requesting to generate when corresponding recommendation results；

It is analyzed based on the reason of target execution information generates target recommendation results to recommender system, obtains recommender system Abnormal cause.

Optionally, the refinement function of program and extension function can refer to and be described below.

Now using data storage device as user behavior inquiry system, the service being applied in Analysis on Abnormal method Device be independently of other than recommendation server another server (for the ease of being distinguished with recommendation server, herein temporarily should The server that Analysis on Abnormal method is applied to is known as Analysis on Abnormal server) for, the embodiment of the present application is mentioned The system composition for the application scenarios that the Analysis on Abnormal method of confession is applicable in is introduced.As shown in figure 4, shown in Fig. 4 It may include: terminal 41, recommendation server 42, user behavior inquiry system 43 and Analysis on Abnormal service in application scenarios Device 44.

Wherein, it is provided with storage unit in the application scenarios, the storage unit is for storing recommendation server response terminal The request of transmission generates the execution information generated when recommendation results.Preferably, each request that recommendation server receives is asked The request mark for the request asked mark to be stored in storage unit, and stored in storage unit is associated with recommender system response should Request generates the execution information generated when recommendation results.For example, recommendation server receives request 1, respond request 1, which generates, recommends Execution information 1 is generated when as a result, in the memory unit the request mark of storage request 1, and the request mark of the request 1 is associated with Execution information 1.

In the embodiment of the present application, it is preferred that the storage unit for storing the request mark for being associated with execution information can To be set in recommendation server or be set in Analysis on Abnormal server, alternatively, executing letter for storing to be associated with The storage unit of the request mark of breath is the data storage device independently of recommendation server and Analysis on Abnormal server, Specific setting position in relation to storage unit, it is not limited here.

Now it is illustrated so that storage unit is set in Analysis on Abnormal server as an example.Whenever recommender system receives One request will respond the request and generate recommendation results, and recommender system can be produced when respond request generates recommendation results The request of the execution information and request mark can be sent to different by raw execution information after recommender system generates execution information Normal analysis of causes server, in order to which Analysis on Abnormal server identifies received request and execution information association Storage.For example, generating execution information 1 when recommender system respond request 1 generates recommendation results, recommender system will request 1 to ask Ask mark and execution information 1 to be sent to Analysis on Abnormal server, Analysis on Abnormal server by request 1 request mark Know and establish incidence relation with execution information 1, and stores the request mark and execution information 1 for establishing the request 1 after incidence relation.

Based on this, the request that recommender system response user is sent by terminal generates recommendation results and feeds back recommendation results While to terminal, the execution information generated when recommender system can be responded to request generation recommendation results is sent to abnormal cause Analysis server, in order in Analysis on Abnormal server storage establish the request mark of the request after incidence relation with And the execution information.Wherein, each execution information being stored in Analysis on Abnormal server can be described as a history and hold Row information, the execution information are associated with request mark, illustrate that the execution information as historical execution information is closed for it Request belonging to the request mark of connection.

As a result, when recommender system responds the destination request generation recommendation results that user is sent by terminal and by recommendation results After feeding back to terminal, if Analysis on Abnormal server determines (the ratio when recommendation results are in the presence of abnormal target recommendation results Such as, user is based on after terminal views the recommendation results, feeds back the recommendation results not to recommender system and is its and interested push away When recommending result, it is believed that determine that the recommendation results are in the presence of abnormal target recommendation results), the Analysis on Abnormal server In the historical execution information for the different requests received for recommender system that can be stored from its storage unit, the mesh is inquired The request of mark request identifies associated historical execution information, and is generated based on the historical execution information inquired to recommender system The reason of target recommendation results, is analyzed, and the abnormal cause of recommender system is obtained.

Further, if the storage unit for being associated with the request mark of execution information for storage is to take independently of recommendation Be engaged in device and Analysis on Abnormal server data storage device (for the ease of with data storage device as shown in Figs. 1-2 It distinguishes, temporarily will be used to store the storage unit for the request mark for being associated with execution information herein as independently of recommendation service The data storage device of device and Analysis on Abnormal server is known as the equipment of analysis storage in real time) for, the embodiment of the present application The system composition for the application scenarios that the Analysis on Abnormal method of offer is applicable in can be found in Fig. 5.As shown in figure 5, shown in Fig. 5 Application scenarios in may include: terminal 51, recommendation server 52, user behavior inquiry system 53, Analysis on Abnormal service Device 54 and the equipment 55 of analysis storage in real time.

Whenever recommender system receives a request, the request will be responded and generate recommendation results, and recommender system exists Respond request generate recommendation results when can generate execution information, when recommender system generate execution information after can by the execution information with The request of the request identifies association, and by the execution information after association and mark is requested to be sent to the equipment of analysis storage in real time, with It is identified and execution information associated storage convenient for analyzing storage equipment in real time for received request.For example, working as recommender system Respond request 1 generates execution information 1 when generating recommendation results, the request for requesting 1 is identified association execution information 1 by recommender system, And the request mark of the request 1 after association and execution information 1 are sent to the equipment of analysis storage in real time, the equipment of analysis storage in real time It will request 1 request mark and 1 associated storage of execution information.

Based on this, the request that recommender system response user is sent by terminal generates recommendation results and feeds back recommendation results While to terminal, recommender system can be responded the request generate recommendation results when generate with the associated execution information of the request Be sent to Analysis on Abnormal server, in order in Analysis on Abnormal server storage establish after incidence relation this ask The request mark asked and the execution information.Wherein, each execution information being stored in Analysis on Abnormal server A referred to as historical execution information, the execution information are associated with request mark, illustrate to execute letter as this of historical execution information Breath is for request belonging to the request mark associated by it.

Correspondingly, when recommender system response user generates recommendation results by the destination request that terminal is sent and will recommend to tie After fruit feeds back to terminal, if Analysis on Abnormal server determines that the recommendation results are in the presence of abnormal target recommendation results, What the Analysis on Abnormal server can be stored from real-time analysis storage equipment asks for the difference that recommender system receives In the historical execution information asked, the request for inquiring the destination request identifies associated historical execution information, and based on inquiring The reason of historical execution information generates the target recommendation results to recommender system is analyzed, and the exception for obtaining recommender system is former Cause.

Further, first recommender system respond request provided by the embodiments of the present application is generated in order to facilitate understanding and recommends knot The process that execution information is generated when fruit describes in detail.In the embodiment of the present application, it is preferred that be by pushing away in recommender system The request of server receiving terminal transmission is recommended, and generates execution information when responding the request and generating recommendation results.

It is illustrated in figure 6 a kind of composed structure schematic diagram of recommendation server provided by the embodiments of the present application.

Referring to Fig. 6, recommendation server in a kind of recommender system provided by the embodiments of the present application by Broker service, Merger services and recalls service and constitutes.

Wherein, recommendation server can be serviced the request for receiving user and sending by terminal, Broker service by Broker It receives current from the user that stores in the user behavior system is inquired in the user behavior system of recommender system after request User behavior information and user's portrait；The user behavior information and user's portrait that Broker service is got are sent to Merger service is handled by Merger service for user behavioural information and user's portrait, the user behavior that obtains that treated Information and user's portrait；Treated user behavior information and user's portrait are sent to the service of recalling by Merger service.

The service of recalling recalls the first recommendation results based on treated user behavior information and user's portrait from resource pool, Also, the service of recalling drawn a portrait based on treated user behavior information and user generate the first recommendation results when, can generate simultaneously It exports and recalls serve log for characterize the treatment process for recalling service；It, can after recalling the first recommendation results of service creation The first recommendation results are fed back to Merger service, Merger service handles the first recommendation results, obtains second and push away Recommend as a result, also, Merger service based on the first recommendation results generate the second recommendation results when, can generate and export for table Levy the Merger serve log of the treatment process of Merger service；Merger is serviced after generating the second recommendation results, can be incited somebody to action Second recommendation results feed back to Broker service, and Broker service can be handled to generate and ask with this second recommendation results It seeks corresponding recommendation results, also, Broker is when generating recommendation results based on the second recommendation results, can generate and export and be used for Characterize the Broker serve log of the treatment process of Broker service.

Wherein, the Broker in recommendation server services the user behavior information and use got from user behavior system Family portrait, recommendation server can also further filter out user from the user behavior information got and request behavioural information, And it requests behavioural information to generate based on the user's portrait got and the user filtered out and exports bandwagon effect log.

In the embodiment of the present application, it is preferred that the user behavior information of user includes sending out respectively with user to recommender system The relevant user of each request sent requests behavioural information, and it includes that recommender system is rung that user relevant to request, which requests behavioural information, Recommendation results should be requested to generate and to after user feedback, feedback information of the user to the recommendation results.

Optionally, from the user behavior information of the user got, the side that user requests behavioural information information is filtered out Formula can be with are as follows: the nearest preset quantity request sent before the request by the user in recommender system is determined, from what is got User relevant to identified each request respectively is filtered out in user behavior information requests behavioural information.Wherein, it sieves herein The user selected requests behavioural information to can be used for generating bandwagon effect log.

In the embodiment of the present application, it is preferred that preset quantity can be 3, if preset quantity is 3, obtain To user user behavior information in, filter out user and request the mode of behavioural information information can be with are as follows: determine recommender system In nearest 3 requests for being sent by the user before the request, filtered out from the user behavior information got respectively with institute The relevant user of determining each request requests behavioural information.It is above only preset quantity provided by the embodiments of the present application Preferred embodiment, the particular content inventor in relation to preset quantity can be arbitrarily arranged according to their own needs, such as by present count Amount is set as any one numerical value in positive integer, such as sets 4 for preset quantity, sets 8 for preset quantity Deng it is not limited here.

It further, can also include Broker service creation in the embodiment of the present application bandwagon effect log generated Recommendation results.

It further, can also include sending for being recommended in the embodiment of the present application bandwagon effect log generated System is responded to generate the basic information of the user of the request of the recommendation results, wherein the basic information of user may include using Family age, gender, address etc..

In the embodiment of the present application, it is preferred that the bandwagon effect log of the recommendation server output of recommender system, Broker Serve log, Merger serve log and serve log is recalled, recommender system respond request can be regarded as and generate recommendation results When the execution information corresponding with the request that generates.

Correspondingly, recommendation server based on corresponding execution information is requested, can establish the request mark of the request and be somebody's turn to do Incidence relation between execution information, and the request mark after incidence relation will be established and execution information is stored to recommendation server Local storage unit.Alternatively, Analysis on Abnormal server can receive the holding corresponding with request of recommendation server transmission Row information, and the request of execution information request corresponding with execution information mark is established into incidence relation, and close establishing Request mark after connection relationship is stored in the storage unit in the Analysis on Abnormal server with execution information, or will establish Request mark and execution information after incidence relation are stored to real-time analysis storage equipment；Wherein, be stored in storage unit/in real time Execution information in analysis storage equipment can regard historical execution information as.

It is above only that the request mark provided by the embodiments of the present application for establishing incidence relation is stored with execution information Preferred embodiment, request mark and the specific storage mode of execution information in relation to establishing incidence relation, inventor can according to from Oneself demand is arbitrarily arranged, it is not limited here.；

In order to enable the generating process of execution information provided by the embodiments of the present application is clearer, provides a kind of recommendation herein System respond request generates the method flow diagram of execution information when generating recommendation results, specifically refer to Fig. 7, in conjunction with Fig. 7 and upper The description to the composed structure of recommendation server is stated, one kind provided by the embodiments of the present application can be held convenient for those skilled in the art The generating process of row information has a better understanding.

As shown in fig. 7, this method comprises:

When S701, acquisition recommender system respond request generate recommendation results, the bandwagon effect log of generation and performance logs；

In the embodiment of the present application, it is preferred that bandwagon effect log is the day that recommender system generates when showing recommendation results Will, performance logs are the log for constituting each service creation of recommender system.Wherein, performance logs include Broker service Log, Merger serve log and recall serve log.

Optionally, user can be responded by the recommendation server in recommender system and is generated by the request that terminal is sent recommend As a result, and generating bandwagon effect log and performance logs when respond request generates recommendation results.

S702, bandwagon effect log and performance logs are handled respectively, the bandwagon effect log that obtains that treated and Performance logs；

Recommendation server, can also be respectively to bandwagon effect log and property after generating bandwagon effect log and performance logs Can log handled, with the bandwagon effect log that obtains that treated and treated performance logs.

Referring to Fig. 8 and Fig. 9, wherein Fig. 8 is to handle respectively bandwagon effect log and performance logs, is handled The application scenarios schematic diagram of bandwagon effect log and performance logs afterwards；Fig. 9 is provided by the embodiments of the present application a kind of right respectively Bandwagon effect log and performance logs are handled, the method flow of obtain that treated bandwagon effect log and performance logs Figure.

As shown in figure 9, this method comprises:

S901, bandwagon effect log and performance logs are stored in message queue；

In the embodiment of the present application, it is preferred that message queue can be kafka message queue.It is above only the application The preferred embodiment for the message queue that embodiment provides, the specific implementation inventor of relevant information queue can be according to oneself need It asks and is configured, it is not limited here.

S902, the first consumption program of starting and the second consumption program, carry out bandwagon effect log by the first consumption program Processing, the bandwagon effect log that obtains that treated, and performance logs are handled by the second consumption program, after obtaining processing Performance logs.

Wherein, the first consumption program may be considered effect log consumption program, the second consumption program property of may be considered It can log consumption program；The bandwagon effect log in kafka message queue is handled based on effect log consumption program, is obtained To treated bandwagon effect log；Based on performance logs consumption program to the performance logs in kafka message queue at Reason, the performance logs that obtain that treated.

Optionally, performance logs include Broker serve log, Merger serve log and recall serve log, performance day Will consumption program handles the performance logs in kafka message queue, the performance logs that obtain that treated, comprising: performance Log consumption program handles the Broker serve log in kafka message queue, the Broker service that obtains that treated Log；Performance logs consumption program handles the Merger serve log in kafka message queue, obtains that treated Merger serve log；Performance logs consumption program handles the serve log of recalling in kafka message queue, obtains everywhere Serve log is recalled after reason.

S703, based on treated bandwagon effect log and performance logs, generate execution information.

Further, for the ease of now mentioning to a kind of understanding of Analysis on Abnormal method provided by the embodiments of the present application For a kind of Analysis on Abnormal method flow diagram, referring to Figure 10.It, can in conjunction with the description of Figure 10 and the description of above-described embodiment It realizes to a kind of detailed description of Analysis on Abnormal method provided by the embodiments of the present application.

Analysis on Abnormal method as shown in Figure 10 provided by the embodiments of the present application can be applied to Analysis on Abnormal Server.

As shown in Figure 10, this method comprises:

S1001, the target recommendation results for determining presence exception in recommender system；

In the embodiment of the present application, there is abnormal target in determining recommender system and recommend knot in Analysis on Abnormal server When fruit (for example, when the recommendation results user that recommender system generates is unsatisfied with), step S1002 can be executed.

S1002, acquisition recommender system generate the destination request responded when target recommendation results；

In the embodiment of the present application, it is preferred that target recommendation results are to receive user to the recommendation results in recommender system Selection operation determine, obtain the destination request that is responded when recommender system generates target recommendation results, comprising: from it is current when Destination request is inquired at least one request that recommender system is sent from user before quarter, it is raw that recommender system responds destination request At recommendation results be target recommendation results.

Based on the above embodiment it is found that recommender system is directed to received each request, can all be inquired in user behavior User relevant to the request is stored in system and requests behavioural information, and it includes user to the recommendation system which, which requests behavioural information, System responds the feedback information for the recommendation results that the request generates.

If it is determined that when there are abnormal target recommendation results in recommender system, can from user behavior inquiry system (this When user behavior system in include each request sent from user to recommender system before current time) inquire target and ask It asks, recommender system, which responds the destination request, can be generated the target recommendation results.

Further, the embodiment of the present application also provide it is a kind of sent before current time from user to recommender system to The method flow diagram of destination request is inquired in a few request, referring to Figure 11.

As shown in figure 11, this method comprises:

S1101, the user identifier for determining user and recommender system generate the time range of target recommendation results；

In the embodiment of the present application, user is sent to recommender system by terminal and is requested, and recommender system respond request generates Recommendation results, and recommendation results are fed back into user, it can be by way of screenshot if user is dissatisfied for the recommendation results Recommendation results are fed back into recommender system, in order to which recommender system generates the unsatisfied recommendation results the reason of progress of user to it Analysis, obtains recommender system abnormal cause.

Wherein, user can directly show terminal after seeing the recommendation results that recommender system is returned by terminal The recommendation results page carries out screenshot, and screenshot is returned to recommender system, would generally include that terminal is being cut in the screenshot The system time of terminal when figure.

Optionally, based on the system time in screenshot, it can determine that recommender system generates the time model of target recommendation results It encloses, for example, preset time float value, the time interval after determining the system time in screenshot, between determining and system time For the first time and the second time of time float value, and will be by the first time and the time range that constitutes of the second time is as institute Determining recommender system generates the time range of target recommendation results.For example, if preset time float value is 3 hours, true Determining the system time in screenshot is 15:00 on May 29th, 2018, it is determined that the time interval between system time is floating for the time The first time and the second time of dynamic value are respectively 18:00 on May 29,12:00 and 2018 year on the 29th May in 2018；Then determine The time range that recommender system generates target recommendation results is 18:00 on May 29,12:00 to 2018 years on the 29th May in 2018.

S1102, inquiry is associated with target user's mark from each request that recommender system before current time receives Request, each request that recommender system receives is associated with the user identifier that oriented recommender system sends the user of request；

S1103, destination request is determined from the request inquired, recommender system responds the recommendation results that destination request generates Time for target recommendation results, and recommender system response destination request generation target recommendation results is located in time range.

Recommender system can receive the request that different users sends, and each request that recommender system receives is associated with Send the user identifier of the user of the request；It is requested when user is sent by terminal to recommender system, and recommender system is fed back Recommendation results it is dissatisfied when (that is, when the recommendation results exception, the recommendation results are temporarily known as target recommendation results), determining should The user identifier and recommender system of user generate the time range of the target recommendation results；(at this time from user behavior system The each request for the transmission of each user that recommender system receives before including current time in user behavior system) in inquiry Associated each request is identified with target user, and determines destination request from the request inquired, wherein recommender system response The target recommendation results can be generated in the destination request, and recommender system responds the destination request and generates the target recommendation results Time be located in identified time range.

S1003, from for different requests historical execution information in, inquiry recommender system response destination request generate target The target execution information generated when recommendation results, recommender system before the historical execution information requested for one is current time Respond request generates the execution information for requesting to generate when corresponding recommendation results；

Further, recommender system is directed to received each request, can not only deposit in user behavior inquiry system User's request behavioural information relevant to the request is stored up, also storage can establish being somebody's turn to do for incidence relation in real-time analysis storage equipment (this is stored in reality to the execution information for requesting mark to generate when responding request generation recommendation binding fruit with recommender system of request When analysis storage equipment in execution information historical execution information when can consider).

It, can after the destination request responded when determining that recommender system generates target recommendation results based on step S1002 Associated historical execution information is identified with the inquiry from real-time analysis storage equipment and the request of the destination request, and by the history Execution information is as the target execution information inquired.

S1004, it analyzes, is pushed away based on the reason of target execution information generates target recommendation results to recommender system Recommend the abnormal cause of system.

In the embodiment of the present application, the reason of recommender system generation target recommendation results, is divided according to target execution information Analysis, obtains the abnormal cause of the recommender system, in order to which the abnormal cause based on recommender system improves recommender system, with So that the request recommendation results generated that improved recommender system response user is sent by terminal are more accurate, more fill it up with The demand of sufficient user, and then user is improved to the viscosity of recommender system.

It is being discussed in detail for Analysis on Abnormal method provided by the embodiments of the present application above, for the ease of art technology Personnel are to a kind of understanding of Analysis on Abnormal method provided by the embodiments of the present application, now in conjunction with concrete application scene to the application A kind of Analysis on Abnormal method that embodiment provides is described in detail.

It is a kind of Analysis on Abnormal schematic diagram of a scenario provided by the embodiments of the present application referring to Figure 12.

In the initial state, default recommendation server does not receive any request；It is subsequent, whenever recommendation server receives One request, just generates recommendation results corresponding with the request and execution information, and the recommendation results of generation are returned to use Family stores the execution information corresponding with request of generation to real-time analysis storage equipment.

For example, recommendation server receives first request, and (request is that user 1 passes through terminal 1 to recommendation referring to Figure 12 The request 1 that server is sent) after, it generates and requests 1 corresponding recommendation results 1 and execution information 1, asking for incidence relation will be established 1 is asked to store with execution information 1 to real-time analysis storage equipment；Recommendation results 1 are returned into terminal 1, request row to generate user For information 1, user's request behavioural information 1 is stored to user behavior inquiry system；

Receiving second request, (request is that user 2 passes through what terminal 2 was sent to recommendation server to recommendation server After request 2), generates and request 2 corresponding recommendation results 2 and execution information 2, by the request 2 for establishing incidence relation and execute letter Breath 2 is stored to real-time analysis storage equipment；Recommendation results 2 are returned into terminal 2, behavioural information 2 is requested to generate user, will use Family request behavioural information 2 is stored to user behavior inquiry system；

Receiving third request, (request is that user 1 passes through what terminal 1 was sent to recommendation server to recommendation server After request 3), generates and request 3 corresponding recommendation results 3 and execution information 3, by the request 3 for establishing incidence relation and execute letter Breath 3 is stored to real-time analysis storage equipment；Recommendation results 3 are returned into terminal 1, behavioural information 3 is requested to generate user, will use Family request behavioural information 3 is stored to user behavior inquiry system；

Receiving the 4th request, (request is that user 3 passes through what terminal 3 was sent to recommendation server to recommendation server After request 4), generates and request 4 corresponding recommendation results 4 and execution information 4, by the request 4 for establishing incidence relation and execute letter Breath 4 is stored to real-time analysis storage equipment；Recommendation results 4 are returned into terminal 3, behavioural information 4 is requested to generate user, will use Family request behavioural information 4 is stored to user behavior inquiry system ...

It is to be noted that user behavior inquiry system requests behavioural information to generation based on the user received by it User's portrait is updated；Each recommendation server is required to look into based on user behavior when respond request generates recommendation results Current user behavior information and user, which draw a portrait, in inquiry system generates recommendation results.

If recommendation server after receiving the 4th request, before the 5th request, receives user 1 and passes through terminal When 1 return recommendation results 3 are abnormal recommendation results, recommendation results 3 can be sent to Analysis on Abnormal by recommendation server Server, the request responded when determining that recommendation server generates recommendation results 3 by Analysis on Abnormal server are request 3, At this point, go to analyze in real time in storage equipment 3 corresponding execution information 3 of inquiry and request (with 3 corresponding execution informations of request for The request of request 3 identifies associated execution information)；To generate abnormal recommendation results to recommendation server based on execution information 3 Reason is analyzed, and recommendation server abnormal cause is obtained, which is recommender system abnormal cause.

By on the basis of concrete application scene to a kind of Analysis on Abnormal method provided by the embodiments of the present application into Row is discussed in detail, so that a kind of Analysis on Abnormal method provided by the embodiments of the present application is clearer, is easy to understand.

Analysis on Abnormal device provided in an embodiment of the present invention is introduced below, hair abnormal cause described below Analytical equipment is regarded as, the server Analysis on Abnormal method that embodiment provides to realize the present invention, the journey of required setting Sequence module.Analysis on Abnormal device content described below, can be mutual with above-described Analysis on Abnormal method content To should refer to.

Figure 13 is a kind of structural block diagram of Analysis on Abnormal device provided by the embodiments of the present application, which can be applied to Analysis on Abnormal server, referring to Fig.1 3, which may include:

Target recommendation results determination unit 131, for determining the target recommendation results of the exception of the presence in recommender system；

Destination request acquiring unit 132 is asked for obtaining the target responded when recommender system generates target recommendation results It asks；

Execution information query unit 133, for from the historical execution information for different requests, inquiry recommender system to be rung The target execution information for answering destination request to generate when generating target recommendation results, the historical execution information requested for one are to work as Recommender system respond request generates the execution information for requesting to generate when corresponding recommendation results before the preceding moment；

Analysis on Abnormal unit 134, for generating target recommendation results to recommender system based on target execution information Reason is analyzed, and the abnormal cause of recommender system is obtained.

In the embodiment of the present application, it is preferred that target recommendation results are to receive user to the recommendation results in recommender system Selection operation determine, destination request acquiring unit is specifically used for: before current time from user to recommender system send At least one request in inquire destination request, it is that target recommends knot that recommender system, which responds the recommendation results that destination request generates, Fruit.

In the embodiment of the present application, it is preferred that destination request acquiring unit, comprising:

Time range determination unit, user identifier and recommender system for determining user generate target recommendation results Time range；

Query unit is requested, for the inquiry from each request that recommender system before current time receives and target use Family identifies associated request, and each request that recommender system receives is associated with the user that oriented recommender system sends the user of request Mark；

Destination request obtains subelement, and for determining destination request from the request inquired, recommender system responds target The recommendation results that request generates are target recommendation results, and recommender system response destination request generates the time of target recommendation results In time range.

In the embodiment of the present application, it is preferred that execution information query unit, comprising:

Historical execution information determination unit, for determining the historical execution information for being directed to different requests；It is requested for one Historical execution information, request request mark with request historical execution information；

Execution information inquires subelement, for inquiring associated request from the historical execution information for different requests It is identified as the historical execution information of the request mark of destination request, and is rung the historical execution information inquired as recommender system The target execution information for answering destination request to generate when generating target recommendation results.

Further, a kind of Analysis on Abnormal device provided by the embodiments of the present application further includes that execution information generates list Member, the execution information generation unit, comprising:

Log acquisition unit, when generating recommendation results for obtaining recommender system respond request, the bandwagon effect day of generation Will and performance logs；Bandwagon effect log is the log that recommender system generates when showing recommendation results, and performance logs are for structure At the log of each service creation of recommender system；

Journal processing unit, for handling respectively bandwagon effect log and performance logs, the exhibition that obtains that treated Show effect log and performance logs；

Execution information generates subelement, for based on treated bandwagon effect log and performance logs, generation to execute letter Breath.

In the embodiment of the present application, it is preferred that journal processing unit, comprising:

Log storage unit, for bandwagon effect log and performance logs to be stored in message queue；

Log processing subelement, for starting the first consumption program and the second consumption program, by the first consumption program to exhibition Show that effect log is handled, the bandwagon effect log that obtains that treated, and performance logs are carried out by the second consumption program Processing, the performance logs that obtain that treated.

In the embodiment of the present application, it is preferred that bandwagon effect log includes user behavior information and user's portrait；Execute letter Generation unit is ceased, when being specifically used for obtaining recommender system respond request generation recommendation results, sends the current use of the user of request Family behavioural information and user's portrait；It is drawn a portrait based on the user behavior information got and user and generates bandwagon effect log.

In the embodiment of the present application, it is preferred that the user behavior information of user includes sending out respectively with user to recommender system The relevant user of each request sent requests behavioural information；Execution information generation unit, is specifically used for: obtaining recommender system response When request generates recommendation results, the user of request current user behavior information and user's portrait are sent；It determines in recommender system It is requested before request and by the nearest preset quantity that user sends；Filtered out from the user behavior information got respectively with The relevant user of identified each request requests behavioural information；Based on the user's portrait got and the user's request filtered out Behavioural information generates bandwagon effect log.

In the embodiment of the present application, it is preferred that each service of recommender system include Broker service, Merger service with And the service of recalling, performance logs include Broker serve log, Merger serve log and recall serve log, in which: are called together Returning serve log is when recalling to service based on recommender system respond request generation recommendation results, to send the current use of the user of request Family behavioural information and user's portrait, when recalling the first recommendation results from resource pool, the place for being used to characterize service of recalling of generation The log of reason process；Merger serve log is that Merger service is handled the first recommendation results to obtain the second recommendation knot When fruit, the log of the treatment process for characterizing Merger service of generation；Broker serve log is Broker service to the When two recommendation results are handled to obtain recommendation results, the log of the treatment process for characterizing Broker service of generation.

Further, the embodiment of the present invention also provides a kind of storage medium, which realizes that exception is former for storing Because of the program of analysis method.

The detailed description of program in relation to storing in storage medium provided by the embodiments of the present application can refer to above-described embodiment, This will not be repeated here.

A kind of Analysis on Abnormal method provided by the invention, server, by for each request storage recommender system The mode that the request generates the execution information generated when recommendation results is responded, so that there are target recommendation results in recommender system When (the target recommendation results are abnormal recommendation results), responded when can generate the target recommendation results based on recommender system Destination request is uniquely determined out from multiple execution informations of storage and is generated when really generating the target recommendation results in history Execution information；And then analyzed based on the reason of the execution information generates the target recommendation results to recommender system, it is pushed away Recommend the abnormal cause of system.

Each embodiment in this specification is described in a progressive manner, the highlights of each of the examples are with other The difference of embodiment, the same or similar parts in each embodiment may refer to each other.For device disclosed in embodiment For, since it is corresponded to the methods disclosed in the examples, so being described relatively simple, related place is said referring to method part It is bright.

Professional further appreciates that, unit described in conjunction with the examples disclosed in the embodiments of the present disclosure And algorithm steps, can be realized with electronic hardware, computer software, or a combination of the two, in order to clearly demonstrate hardware and The interchangeability of software generally describes each exemplary composition and step according to function in the above description.These Function is implemented in hardware or software actually, the specific application and design constraint depending on technical solution.Profession Technical staff can use different methods to achieve the described function each specific application, but this realization is not answered Think beyond the scope of this invention.

The step of method described in conjunction with the examples disclosed in this document or algorithm, can directly be held with hardware, processor The combination of capable software module or the two is implemented.Software module can be placed in random access memory (RAM), memory, read-only deposit Reservoir (ROM), electrically programmable ROM, electrically erasable ROM, register, hard disk, moveable magnetic disc, CD-ROM or technology In any other form of storage medium well known in field.

The foregoing description of the disclosed embodiments enables those skilled in the art to implement or use the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, as defined herein General Principle can be realized in other embodiments in the case where not departing from core of the invention thought or scope.Therefore, originally Invention is not intended to be limited to the embodiments shown herein, and is to fit to and the principles and novel features disclosed herein Consistent widest scope.

Claims

1. a kind of Analysis on Abnormal method characterized by comprising

From the historical execution information for different requests, inquires the recommender system and respond the destination request generation mesh The target execution information generated when marking recommendation results, the historical execution information requested for one are that the current time foregoing description pushes away The system of recommending responds the request and generates the execution information for requesting to generate when corresponding recommendation results；

It analyzes, obtains based on the reason of the target execution information generates the target recommendation results to the recommender system The abnormal cause of the recommender system.

2. the method according to claim 1, wherein the target recommendation results are to receive user to recommender system In the selection operations of recommendation results determine that described obtain when the recommender system generates the target recommendation results responded Destination request, comprising:

Destination request is inquired at least one request that the recommender system is sent from the user from before current time, institute Stating recommender system and responding the recommendation results that the destination request generates is the target recommendation results.

3. according to the method described in claim 2, it is characterized in that, described push away before current time from the user to described It recommends at least one request of system transmission and inquires destination request, comprising:

Inquiry identifies associated with the target user from each request that current time foregoing description recommender system receives Request, the recommender system each of receive the request and are associated with the user that the oriented recommender system sends the request User identifier；

Destination request is determined from the request inquired, the recommender system responds the recommendation results that the destination request generates and is The target recommendation results, and the recommender system responds the destination request and generates time of the target recommendation results and is located at In the time range.

4. the method according to claim 1, wherein it is described from for different requests historical execution information in, It inquires the recommender system and responds the destination request and generate the target execution information generated when the target recommendation results, wrap It includes:

Determine the historical execution information for being directed to different requests；The historical execution information requested for one, the request of the request The historical execution information of mark and the request；

From the historical execution information for different requests, the associated request mark for requesting to be identified as the destination request is inquired Historical execution information, and it is raw using the historical execution information inquired as the recommender system to respond the destination request The target execution information generated when at the target recommendation results.

5. the method according to claim 1, wherein the recommender system, which responds the request, generates the recommendation The process of generation execution information includes: when as a result

When obtaining the recommender system and responding the request generation recommendation results, the bandwagon effect log of generation and performance day Will；The bandwagon effect log is the log that the recommender system generates when showing the recommendation results, and the performance logs are For constituting the log of each service creation of the recommender system；

The bandwagon effect log and the performance logs are handled respectively, obtain that treated bandwagon effect log and property It can log；

6. according to the method described in claim 5, it is characterized in that, described respectively to the bandwagon effect log and the performance Log is handled, obtain that treated bandwagon effect log and performance logs, comprising:

The bandwagon effect log and performance logs are stored in message queue；

The first consumption program of starting and the second consumption program, carry out the bandwagon effect log by the first consumption program Reason, the bandwagon effect log that obtains that treated, and the performance logs are handled by the second consumption program, The performance logs that obtain that treated.

7. according to the method described in claim 5, it is characterized in that, the bandwagon effect log includes user behavior information and use Family portrait, the recommender system respond the process packet that the bandwagon effect log is generated when the request generates the recommendation results It includes:

When obtaining the recommender system and responding the request generation recommendation results, the current use of the user of the request is sent Family behavioural information and user's portrait；

8. the method according to the description of claim 7 is characterized in that the user behavior information of user include respectively with the user The relevant user of each request sent to the recommender system requests behavioural information, this method further include:

User relevant to identified each request respectively is filtered out from the user behavior information got requests behavior letter Breath；

Described drawn a portrait based on the user behavior information got and user generates the bandwagon effect log, comprising: based on acquisition To user's portrait and the user that filters out request behavioural information to generate the bandwagon effect log.

9. according to the method described in claim 5, it is characterized in that, each service of the recommender system includes Broker clothes Business, Merger service and the service of recalling, the performance logs include Broker serve log, Merger serve log and call together Return serve log, in which:

The serve log of recalling is the service of recalling based on the recommender system response request generation recommendation knot When fruit, the user of the request current user behavior information and user's portrait are sent, the first recommendation knot is recalled from resource pool When fruit, the log for being used to characterize the treatment process for recalling service of generation；

The Merger serve log is that the Merger service is handled first recommendation results to obtain second and push away When recommending result, the log of the treatment process for characterizing the Merger service of generation；

The Broker serve log is that the Broker service is handled second recommendation results to obtain described push away When recommending result, the log of the treatment process for characterizing the Broker service of generation.

10. a kind of Analysis on Abnormal device characterized by comprising

Destination request acquiring unit is asked for obtaining the target responded when the recommender system generates the target recommendation results It asks；

Execution information query unit, for from the historical execution information for different requests, inquiring the recommender system response The destination request generates the target execution information generated when the target recommendation results, and the history requested for one executes letter What breath generated when responding the request generation corresponding recommendation results of the request for current time foregoing description recommender system holds Row information；

Analysis on Abnormal unit recommends knot for generating the target to the recommender system based on the target execution information The reason of fruit, is analyzed, and the abnormal cause of the recommender system is obtained.

11. a kind of server characterized by comprising at least one processor and at least one processor；The memory is deposited Program is contained, the processor calls the program of the memory storage, and described program is any for realizing such as claim 1-9 Analysis on Abnormal method described in one.