CN110020102A - Object recommendation method, apparatus, storage medium, processor and system - Google Patents

Abstract

The invention discloses a kind of object recommendation method, apparatus, storage medium, processor and systems.Wherein, this method comprises: characteristic factor according to user preference, determines the Generalization bounds for carrying out object recommendation to user；According to the Generalization bounds determined to user's recommended.The present invention is solved when being caused using a Generalization bounds to user's recommended due to all users, the not high technical problem of the individualized experience of user.

Description

Object recommendation method, apparatus, storage medium, processor and system

Technical field

The present invention relates to data processing fields, in particular to a kind of object recommendation method, apparatus, storage medium, place Manage device and system.

Background technique

Currently, when to user's recommended, it can be using cloud server to user's recommended, or use third Method from the recommended engine just provided to user's recommended, for example, cloud server is to user's recommendation service or third party The recommended engine of offer is to user's recommendation service.The above method is usually to be directed to specific case to carry out data analysis, is divided Analysis is as a result, the direction for finding optimization further according to analysis result is attempted to constantly be attempted as a result, last basis It attempts result and is further improved recommendation results recommended to the user.

Due to needing to find the direction of optimization based on the analysis results, and it also requires constantly being attempted, this can not The result in the direction, trial that guarantee every suboptimization can reach optimal recommendation effect.In the direction of optimization, the result of trial In the case where optimal recommendation effect is not achieved, it may also need to change other optimization direction, carry out another trial, from And lead to the low efficiency to user's recommended.In order to achieve the purpose that user's recommended, it usually needs analysis personnel's tool Have experience abundant, to determine reasonable optimization direction, determine it is reasonable attempt method, and most of use recommended engine product Analysis personnel extremely lack this aspect experience, often do not know wherefrom start in practical applications.Thus, even The prioritization scheme of very simple recommendation results, due to analysis, personnel lack experience, and cause to imitate to the recommendation of user's recommended Fruit is bad.

Traditional optimizes recommendation results using manual to the method for user's recommended, needs to analyze personnel to recommendation results The method analyzed, and recommendation results are analyzed be it is diversified, need the experience of comprehensive analysis personnel and right The susceptibility of data optimizes recommendation results.Since the above method is highly dependent upon the experience of analysis personnel and to data Susceptibility, so as to the inefficiency of user's recommended, and it is also very rare to analyze personnel, so that traditional to user The method of recommended is difficult to realize.And it is often all unsatisfactory to the not optimized algorithm of recommendation results, it seriously affects Experience of the user for recommendation effect.

Thus, for all users, a Generalization bounds are often all used, which can be by data point Analysis, finds the direction of optimization, to constantly be attempted, improves the Generalization bounds that recommendation results obtain, does not ensure that this Generalization bounds can bring optimum experience for all users, and when causing to user's recommended, the individualized experience of user is not High problem.

For it is above-mentioned caused using a Generalization bounds to user's recommended due to all users when, of user Propertyization experiences not high problem, and currently no effective solution has been proposed.

Summary of the invention

The embodiment of the invention provides a kind of object recommendation method, apparatus, storage medium, processor and systems, at least When solving to cause using a Generalization bounds to user's recommended due to all users, the individualized experience of user is not high The technical issues of.

According to an aspect of an embodiment of the present invention, a kind of object recommendation method is provided.This method comprises: according to user The characteristic factor of preference determines the Generalization bounds for carrying out object recommendation to user；According to the Generalization bounds determined to user Recommended.

According to an aspect of an embodiment of the present invention, a kind of object recommendation method is additionally provided.It is used this method comprises: receiving In the input information of retrieval；Determine the user information of the corresponding user of input information；According to user information, obtain in advance and user The corresponding Generalization bounds of information, wherein Generalization bounds are used for the characteristic factor according to user preference, carry out object to user and push away It recommends；According to the Generalization bounds of acquisition to user's recommended.

According to an aspect of an embodiment of the present invention, a kind of object recommendation device is additionally provided.The device includes: first true Cover half block determines the Generalization bounds for carrying out object recommendation to user for the characteristic factor according to user preference；First pushes away Recommend module, for according to the Generalization bounds determined to user's recommended.

According to an aspect of an embodiment of the present invention, a kind of object recommendation device is additionally provided.The device includes: reception mould Block, for receiving the input information for being used for retrieval；Second determining module, for determining user's letter of the corresponding user of input information Breath；Module is obtained, for obtaining Generalization bounds corresponding with user information in advance according to user information, wherein Generalization bounds are used In the characteristic factor according to user preference, object recommendation is carried out to user；Second recommending module, for the recommendation plan according to acquisition Slightly to user's recommended.

According to an aspect of an embodiment of the present invention, a kind of storage medium is additionally provided.The storage medium includes storage Program, wherein equipment where control storage medium executes the object recommendation method of the embodiment of the present invention in program operation.

According to an aspect of an embodiment of the present invention, a kind of processor is additionally provided.The processor is used to run program, In, the object recommendation method of the embodiment of the present invention is executed when program is run.

According to an aspect of an embodiment of the present invention, a kind of system is additionally provided.The system includes: processor；And it deposits Reservoir is connect with processor, for providing the instruction for handling following processing step for processor: step 1, according to user preference Characteristic factor determines the Generalization bounds for carrying out object recommendation to user；Step 2, according to the Generalization bounds determined to user Recommended.

In embodiments of the present invention, it according to the characteristic factor of user preference, determines for carrying out object recommendation to user Generalization bounds；According to the Generalization bounds determined to user's recommended.Due to that can be determined according to the characteristic factor of user preference Personalized Recommendation Strategy comes to user's recommended, and client is without writing code, to realize raising to user's recommendation pair The technical effect of the efficiency of elephant, so solve due to all users using a Generalization bounds and cause to user's recommended When, the individualized experience of user it is not high the technical issues of.

Detailed description of the invention

The drawings described herein are used to provide a further understanding of the present invention, constitutes part of this application, this hair Bright illustrative embodiments and their description are used to explain the present invention, and are not constituted improper limitations of the present invention.In the accompanying drawings:

Fig. 1 is a kind of schematic diagram of system according to an embodiment of the present invention；

Fig. 2 is the schematic diagram of another system according to an embodiment of the present invention；

Fig. 3 is according to an embodiment of the present invention a kind of for carrying out the component diagram of object recommendation；

Fig. 4 be it is according to an embodiment of the present invention it is a kind of for realizing object recommendation method terminal (or movement set It is standby) hardware block diagram；

Fig. 5 is a kind of flow chart of object recommendation method according to an embodiment of the present invention；

Fig. 6 is the flow chart of another object recommendation method according to an embodiment of the present invention；

Fig. 7 is the flow chart of another object recommendation method according to an embodiment of the present invention；

Fig. 8 is a kind of schematic diagram of object recommendation device according to an embodiment of the present invention；

Fig. 9 is the schematic diagram of another object recommendation device according to an embodiment of the present invention；And

Figure 10 is a kind of structural block diagram of terminal according to an embodiment of the present invention.

Specific embodiment

In order to enable those skilled in the art to better understand the solution of the present invention, below in conjunction in the embodiment of the present invention Attached drawing, technical scheme in the embodiment of the invention is clearly and completely described, it is clear that described embodiment is only The embodiment of a part of the invention, instead of all the embodiments.Based on the embodiments of the present invention, ordinary skill people The model that the present invention protects all should belong in member's every other embodiment obtained without making creative work It encloses.

It should be noted that description and claims of this specification and term " first " in above-mentioned attached drawing, " Two " etc. be to be used to distinguish similar objects, without being used to describe a particular order or precedence order.It should be understood that using in this way Data be interchangeable under appropriate circumstances, so as to the embodiment of the present invention described herein can in addition to illustrating herein or Sequence other than those of description is implemented.In addition, term " includes " and " having " and their any deformation, it is intended that cover Cover it is non-exclusive include, for example, the process, method, system, product or equipment for containing a series of steps or units are not necessarily limited to Step or unit those of is clearly listed, but may include be not clearly listed or for these process, methods, product Or other step or units that equipment is intrinsic.

Embodiment 1

The embodiment of the invention provides a kind of systems.

Fig. 1 is a kind of schematic diagram of system according to an embodiment of the present invention.As shown in Figure 1, the system 100 includes processor 102 and memory 104.

Processor 102.

Memory 104 is connect with processor 102, for providing the instruction for handling following processing step for processor 102: Step 1, according to the characteristic factor of user preference, the Generalization bounds for carrying out object recommendation to user are determined；Step 2, foundation Determining Generalization bounds are to user's recommended.

The processor 102 of the embodiment is connected with memory 104, for receive the offer of memory 104 for foundation The characteristic factor of user preference, determine for user carry out object recommendation Generalization bounds, according to determine Generalization bounds to The instruction of user's recommended.

In this embodiment, user preference, that is, the preference of user, can consider object (for example, commodity for user And service) when made rationality have tendentious selection, be the economics of user cognition, psychological feelings and rationality The synthesis result of tradeoff.Processor 102 can be by obtaining user behavior, and is analyzed to obtain user to user behavior inclined Good, which can be the user behavior of predetermined quantity.Wherein, processor 102 is available when obtaining user behavior The historical behavior of user can be with for example, obtain the historical log information of user, history access information, historical operation behavior etc. The uniform resource locator (Uniform Resource Locator, referred to as URL) of user's access is carried out parsing determining use The historical behavior at family obtains the historical behavior of user by the answer result of user's answer, can also be by user in net Some scorings, recommendation information etc. come historical behavior for collecting user etc., herein with no restrictions.

User preference corresponds to one or more features factor.Processor 102 is based on after the historical behavior for obtaining user The historical behavior of user determines the one or more features factor of user preference.Optionally, the characteristic factor of the user preference is It may include preference factor of the user for recommended object for determining the factor of user preference, which specifically may be used Think the factor for influencing user preference, may include a variety of preference factors, for example, including following one:

User is for being recommended the preference factor of the real-time of object, that is, the real-time influence user recommended is for quilt The preference of recommended, for example, user compares the real-time for paying close attention to recommended object；User is for being recommended the content of object The preference factor of accuracy, that is, the accuracy for the content recommended influences user for the preference of recommended object, in other words It says, user compares whether concern recommended is that oneself is really desired；Preference of the user for the recommendation period recommended Factor can be detected that is, that is recommended recommends the period to influence the preference of user according to different periods；User for The preference factor for the network environment recommended, that is, the preference of the network environment influence user recommended, it can be according to not It is detected with network environment；User for be recommended the multifarious preference factor of object, that is, diversity influence user for The preference of recommended object, can indicate diversity level by concentration degree, wherein concentration degree is lower, and diversity level is higher, By concentrating the measurement foundation of index expression concentration degree, value can be time, item contents etc.；User is for being recommended object The preference factor of evaluation quality, that is, evaluation quality influences user for the preference of recommended object, having evaluated bad determinations can To be determined by the scoring of behavior concentration degree, the scoring of content concentration degree, for example, concentration degree scoring, the scoring of content concentration degree are got over Height, then it is better to evaluate, and concentration degree scoring, the scoring of content concentration degree are lower, then evaluates worse；There is recommended object in user It is associated with the preference factor of mating recommendation, that is, user is for being recommended object when recommended object exists and is associated with mating recommendation Preference.

Processor 102 can determine the corresponding preference dimension of above-mentioned preference factor by default probe algorithm, wherein a variety of The characteristic factor of user preference is divided into different dimensions by preference factor, and every kind of preference factor corresponds to a kind of dimension.For example, with Family can correspond to the preference dimension of real-time, user is for recommended for the preference factor of the real-time of recommended object The preference factor of the accuracy of the content of object can correspond to preference dimension of accuracy etc., herein without limitation.

Processor 102 is also right when determining the one or more features factor of user preference in the historical behavior based on user The above-mentioned historical behavior of user's predetermined quantity and corresponding features described above factor are trained, and obtain preference training pattern.It should Preference training pattern includes the algorithm policy write in advance, that is, probe algorithm, can carry out the historical behavior of user pre- Survey, obtain include the one or more features factor of user preference prediction result.Optionally, preference training pattern is detecting When the corresponding preference dimension of one or more features factor, environmental context is distinguished, for example, detected according to different periods, It is detected according to heterogeneous networks.Processor 102 can be inputted the log of the historical behavior of predetermined quantity as various algorithms, In conjunction with collaborative filtering, Factorization scheduling algorithm strategy, to obtain the prediction result of different dimensions, which can carry use Family information, the user information can be used for characterizing user to the preference size of each characteristic factor, with Generalization bounds phase It is corresponding, to realize offline matching.

Processor 102 after the historical behavior based on user, the one or more features factor for determining user preference, according to According to the one or more features factor of determining user preference, the Generalization bounds for carrying out object recommendation to user are determined.Its In, Generalization bounds are used for user's recommended, by Generalization bounds object recommended to the user, that is, recommended object, no Same user corresponds to different Generalization bounds.

Optionally, processor 102, which is based on the formulation of each characteristic factor, recommendation mechanisms, each characteristic factor can be made Fixed different recommendation mechanisms, that is, can have one group of recommendation mechanisms, which can be proposed algorithm, that is, can be with With one group of algorithm.Each recommendation mechanisms are used to pay close attention to some side of user, for example, to each side of item contents. Different characteristic factors corresponds to different recommendation mechanisms.Processor 102 is special in the one or more according to the user preference determined Sign factor, when determining the Generalization bounds for carrying out object recommendation to user, processor 102 determines each characteristic factor respectively The weight of corresponding recommendation mechanisms, that is, defining weight for the corresponding proposed algorithm of each characteristic factor, then according to each The weight of the corresponding recommendation mechanisms of characteristic factor and each recommendation mechanisms determines the recommendation for carrying out object recommendation to user Strategy.Processor 102 can obtain the first recommendation mechanisms formulated based on each characteristic factor or the second recommendation machine respectively System, wherein the first recommendation mechanisms are characterized the recommendation machine that the independence between factor and other feature factor is higher than predetermined threshold System, for example, the first recommendation mechanisms are characterized factor and the mutual independent or approximately independent recommendation mechanisms of other feature factor, this Sample is high-efficient to user's recommended；The independence that second recommendation mechanisms are characterized between factor and other feature factor is lower than The recommendation mechanisms of predetermined threshold are general algorithm, for example, the second recommendation mechanisms are characterized between factor and other feature factor Correlation with higher, the in this way low efficiency to user's recommended, wherein predetermined threshold is for characteristic feature factor and its Its characteristic factor is associated.

Processor 102 is obtaining the first recommendation mechanisms formulated based on each characteristic factor or the second recommendation machine respectively After system, the weight of the corresponding recommendation mechanisms of each characteristic factor is determined respectively.Each characteristic factor and recommendation mechanisms Weight is corresponding, and the weight of the recommendation mechanisms is used to determine recommendation mechanisms in the Generalization bounds for carrying out object recommendation to user In ratio, can be indicated in the form of scoring.For example, when characteristic factor is that user is multifarious for recommended object When preference factor, that is, obtaining the scoring of behavior concentration degree, wherein concentration degree is used for when characteristic factor is behavior concentration degree Indicate the diversity level of the result of algorithm output, concentration degree is lower, and diversity level is higher, and the measurement foundation of concentration degree can With by concentration index expression, value can be time, item contents etc.；When characteristic factor is user in recommended object When the preference factor of the accuracy of appearance, that is, obtaining the scoring of content concentration degree, herein when characteristic factor is content concentration degree With no restrictions.

Optionally, above-mentioned prediction result can carry user information.In processor 102 using the preference pattern pair constructed in advance The historical behavior of user is predicted, after obtaining prediction result, is verified, is analyzed to prediction result.It is every determining respectively When the weight of the corresponding recommendation mechanisms of one characteristic factor, it is used to characterize user to each entrained by prediction result The user information of the preference size of characteristic factor, determines the weight of the corresponding recommendation mechanisms of each characteristic factor respectively, For example, the preference that user information is used to characterize the preference factor of the accuracy of content of the user for being recommended object is big It is small；User information is used to characterize user for the preference size of the preference factor for the recommendation period recommended；User's letter Breath is for characterizing user for the preference size of the preference factor for the network environment recommended；User information is for characterizing Preference size of the user for the recommended multifarious preference factor of object；User information is for characterizing user for being pushed away Recommend the preference size of the preference factor of subject evaluation quality；There is recommended object for characterizing user in user information It is associated with the preference size of the preference factor of mating recommendation.

Optionally, the preference of features described above factor is bigger, and the weight of recommendation mechanisms corresponding with this feature factor is just Bigger, the preference of characteristic factor is just smaller, and this feature factor with the weight of corresponding recommendation mechanisms with regard to smaller, feature because The preference of element is bigger.Conversion of the user in the prediction result of different characteristic factor can be analyzed according to operational indicator Performance, to obtain more user informations.User information can be drawn a portrait by the scene under one group of difference scene and be formed, each feelings Scape portrait may include the scoring of behavior concentration degree, the scoring of content concentration degree, the scoring of content-preference degree etc., wherein content-preference degree Scoring is the individual scoring for the progress of the content of recommended object, herein with no restrictions.

Optionally, processor 102 is after the weight for determining the corresponding recommendation mechanisms of each characteristic factor respectively, foundation The weight of the corresponding recommendation mechanisms of each characteristic factor and each recommendation mechanisms is determined for carrying out object recommendation to user Generalization bounds.Different user has different Generalization bounds, and processor 102 can be apparent to diversity of values on different dimensions User carries out selection algorithm strategy according to scoring and recommendation results, user unconspicuous for diversity of values on different dimensions pushes away Recommending in strategy has the function of reinforcing preference detection.

Some side of the corresponding recommendation mechanisms concern user of each characteristic factor of the embodiment.It is automatic excellent in progress When change, processor 102 is one group of recommended candidate of each user's output first with the corresponding recommendation mechanisms of each characteristic factor Then the performance in each recommended candidate pond is analyzed in pond by off-line verification, so as to further appreciate that each user's is some Hiding user information finally determines the Generalization bounds for being directed to each user using these user informations.

Processor 102 is after determining the Generalization bounds for carrying out object recommendation to user, according to the recommendation plan determined Slightly to user's recommended.Optionally, the corresponding recommendation results of each recommendation mechanisms have corresponding display area, and different pushes away It recommends result and corresponds to different size of display area.Processor 102 when according to the Generalization bounds determined to user's recommended, according to According to the weight of each recommendation mechanisms, the size of the display area of the corresponding recommendation results of each recommendation mechanisms is determined, according to each The size of the display area of the corresponding recommendation results of recommendation mechanisms can divide the corresponding recommendation results of each recommendation mechanisms Area is shown.

The recommendation results of different user and algorithm difference.Optionally, processor 102 according to determining Generalization bounds to When the recommended of family, according to the weight of the corresponding recommendation mechanisms of each of Generalization bounds characteristic factor specific choice for use The recommendation results and algorithm at family, for example, being directed to the recommendation results of user for the concentration degree of Generalization bounds, preference scoring selection And algorithm.In actual operation, it can be applicable in the proposed algorithm of user according to subscriber policy selection, then united to all users Meter selects optimal algorithm (Top algorithm) to execute calculating to the Generalization bounds of all users, calculated result is obtained, further according to calculating As a result it combines to the Generalization bounds of user's progress object recommendation to user's recommended, so as to optimize in great simplification Guarantee certain effect of optimization while journey, can satisfy the optimization demand of naive user, enhances this crowd of user and recommendation is drawn The confidence held up thereby reduces the turnover rate of client.

The system of above-described embodiment can be using in the product, when the step for being integrated with above-mentioned processor 102 in product and handling After rapid, client can very easily carry out fool result optimizing.The result of final optimization pass can be selected for different users Different Generalization bounds are selected, avoid all users using a Generalization bounds, and lead to the low efficiency to user's recommended The problem of, the technical effect improved to the efficiency of user's recommended is realized, and then solve since all users are using one A Generalization bounds and when causing to user's recommended, the individualized experience of user it is not high the technical issues of.

Fig. 2 is the schematic diagram of another system according to an embodiment of the present invention.As shown in Fig. 2, the system 200 includes processing Device 202 and memory 204.

Processor 202.

Memory 204 is connect with processor 202, for providing the instruction for handling following processing step for processor 202: Receive the input information for retrieval；Determine the user information of the corresponding user of input information；According to user information, obtain preparatory Generalization bounds corresponding with user information, wherein Generalization bounds are used for the characteristic factor according to user preference, carry out pair to user As recommending；According to the Generalization bounds of acquisition to user's recommended.

The processor 202 of the embodiment is connected with memory 204, for receive the offer of memory 204 for according to connecing Receive the input information for retrieval；Determine the user information of the corresponding user of input information；According to user information, obtain in advance with The corresponding Generalization bounds of user information, wherein Generalization bounds are used for the characteristic factor according to user preference, carry out object to user Recommend；Instruction according to from the Generalization bounds of acquisition to user's recommended.

The system of the embodiment is used to execute the method interacted with user.

Processor 202 receives the input information for retrieval, which is the information for retrieval, can be by user Input information of the historical behavior log as various recommendation mechanisms, the input information is corresponding with the user information of user.? After processor 202 receives the input information for retrieval, the user information of the corresponding user of input information, user letter are determined Breath can be used for characterizing user to the preference size of each characteristic factor, can be drawn by the scene under one group of difference scene Picture composition, each scene portrait may include behavior concentration degree scores, content concentration degree scores, content-preference degree scores etc., In, the scoring of content-preference degree can be the individual scoring of the progress of the content for recommended object, herein with no restrictions.

The corresponding Generalization bounds for carrying out object recommendation to user of different user are different.It is defeated in the determination of processor 202 After the user information for entering the corresponding user of information, according to user information, Generalization bounds corresponding with user information in advance are obtained, The Generalization bounds are used for the characteristic factor according to user preference, carry out object recommendation to user.

In processor 202 according to user information, when obtaining Generalization bounds corresponding with user information in advance, processor 202 The historical behavior of available user, for example, obtaining historical log information, the history access information, historical operation behavior of user Deng the uniform resource locator that can also be accessed user parse the historical behavior of determining user, pass through user's answer Answer result obtains the historical behavior of user, can also be collected by user some scorings, recommendation information in net etc. The historical behavior etc. of user, herein with no restrictions.

Processor 202 determines the one or more features factor of user preference after the historical behavior for obtaining user.With Family preference corresponds to one or more features factor.Optionally, the characteristic factor of the user preference may include user for being pushed away The preference factor for recommending object may include a variety of preference factors, for example, include user for be recommended object real-time it is inclined Good factor；Preference factor of the user for the accuracy of the content of recommended object；User is for recommendation period for being recommended Preference factor；User is for the preference factor for the network environment recommended；User is multifarious for recommended object partially Good factor；Preference factor of the user for recommended subject evaluation quality；User, which has recommended object, is associated with mating push away One of preference factor recommended, herein with no restrictions.

Processor 202 is in the historical behavior based on user, can be with when determining the one or more features factor of user preference Use various ways, for example, can be by the way of preference training pattern: above-mentioned historical behavior to user's predetermined quantity and Corresponding features described above factor is trained, and obtains preference training pattern.The preference training pattern includes the calculation write in advance Method strategy can predict the historical behavior of user, obtain include user preference one or more features factor Prediction result.Preference training pattern distinguishes environmental context in the corresponding preference dimension of detection one or more features factor, For example, being detected according to different periods, detected according to heterogeneous networks.Processor 202 can be by the history of predetermined quantity The log of behavior is inputted as various algorithms, in conjunction with collaborative filtering, Factorization scheduling algorithm strategy, to obtain different dimensions Prediction result, the prediction result can carry user information, which can be used for characterizing user to each characteristic factor Preference size, it is corresponding with Generalization bounds, to realize offline matching.

Processor 202 after determining the one or more features factor of user preference, determine respectively each feature because The weight of the corresponding recommendation mechanisms of element, that is, defining weight for the corresponding proposed algorithm of each characteristic factor.Processor 202 The first recommendation mechanisms or the second recommendation mechanisms formulated based on each characteristic factor, the first recommendation mechanisms can be obtained respectively The independence being characterized between factor and other feature factor is higher than the recommendation mechanisms of predetermined threshold, for example, the first recommendation mechanisms It is characterized factor and the mutual independent or approximately independent recommendation mechanisms of other feature factor, the in this way effect to user's recommended Rate is high；Second recommendation mechanisms are characterized the recommendation mechanisms that the independence between factor and other feature factor is lower than predetermined threshold, For general algorithm, for example, the second recommendation mechanisms are characterized correlation with higher between factor and other feature factor, this Low efficiency of the sample to user's recommended, wherein predetermined threshold is associated with other feature factor for characteristic feature factor.

Processor 202 is obtaining the first recommendation mechanisms formulated based on each characteristic factor or the second recommendation machine respectively After system, the weight of the corresponding recommendation mechanisms of each characteristic factor is determined respectively.Each characteristic factor and recommendation mechanisms Weight is corresponding, and the weight of the recommendation mechanisms is used to determine recommendation mechanisms in the Generalization bounds for carrying out object recommendation to user In ratio, can be indicated in the form of scoring.Optionally, when characteristic factor is user for being recommended object diversity Preference factor when, that is, when characteristic factor be behavior concentration degree when, obtain behavior concentration degree scoring；When characteristic factor is User for be recommended object content accuracy preference factor when, that is, when characteristic factor be content concentration degree when, obtain Content concentration degree is taken to score, herein with no restrictions.

Processor 202 is after the weight for determining the corresponding recommendation mechanisms of each characteristic factor respectively, according to each The weight of the corresponding recommendation mechanisms of characteristic factor and each recommendation mechanisms determines the recommendation for carrying out object recommendation to user Strategy.Optionally, processor 202 has the user of notable difference to the scoring on different dimensions, carries out selection calculation according to scoring Method strategy and recommendation results, processor 202 reinforce preference in Generalization bounds to the unconspicuous user of diversity of values on different dimensions The function of detection.

Some side of the corresponding recommendation mechanisms concern user of each characteristic factor of the embodiment.It is automatic excellent in progress When change, processor 202 is one group of recommended candidate of each user's output first with the corresponding recommendation mechanisms of each characteristic factor Then the performance in each recommended candidate pond is analyzed in pond by off-line verification, so as to further appreciate that each user's is some Hiding user information finally determines the Generalization bounds for being directed to each user using these user informations.

Processor 202 is after determining the Generalization bounds for carrying out object recommendation to user, according to the recommendation plan determined Slightly to user's recommended.Optionally, the corresponding recommendation results of each recommendation mechanisms have corresponding display area, and different pushes away It recommends result and corresponds to different size of display area.Processor 202 when according to the Generalization bounds determined to user's recommended, according to According to the weight of each recommendation mechanisms, the size of the display area of the corresponding recommendation results of each recommendation mechanisms is determined, according to each The size of the display area of the corresponding recommendation results of recommendation mechanisms can divide the corresponding recommendation results of each recommendation mechanisms Area is shown.

The recommendation results of different user and algorithm difference.Optionally, processor 202 according to determining Generalization bounds to When the recommended of family, according to the weight of the corresponding recommendation mechanisms of each of Generalization bounds characteristic factor specific choice for use The recommendation results and algorithm at family, for example, being directed to the recommendation results of user for the concentration degree of Generalization bounds, preference scoring selection And algorithm.In actual operation, it can be applicable in the proposed algorithm of user according to subscriber policy selection, then united to all users Meter, select optimal algorithm calculating is executed to the Generalization bounds of all users, obtain calculated result, further according to calculated result in conjunction with to User carries out the Generalization bounds of object recommendation to user's recommended, so as to protect while great simplified optimization process Certain effect of optimization is demonstrate,proved, can satisfy the optimization demand of naive user, enhances this crowd of user to the confidence of recommended engine, into And reduce the turnover rate of client.

The system of above-described embodiment can be using in the product, when the step for being integrated with above-mentioned processor 202 in product and handling After rapid, client can very easily carry out fool result optimizing.The result of final optimization pass can be selected for different users Different Generalization bounds are selected, avoid all users using a Generalization bounds, and lead to the low efficiency to user's recommended The problem of, the technical effect improved to the efficiency of user's recommended is realized, and then solve since all users are using one A Generalization bounds and when causing to user's recommended, the individualized experience of user it is not high the technical issues of.

It is illustrated below with reference to scheme of the preferred embodiment to the embodiment of the present invention.

Fig. 3 is according to an embodiment of the present invention a kind of for carrying out the component diagram of object recommendation.As shown in figure 3, should The component 300 of embodiment object recommendation includes: classification component 302, matching component 304, analytic unit 306, formulates 308 and of component Executive module 310.

Classification component 302, for detecting the probe algorithm of user preference, and according to the dimension of user preference to proposed algorithm Classify, can be classified according to probe algorithm to proposed algorithm.Wherein, probe algorithm is the algorithm write in advance. Probe algorithm detects following preference dimension, user for real-time item preference, user for item contents preference, herein not It is limited.Probe algorithm is also in detection preference dimension time zone time-sharing environment context, for example, detecting according to different periods, according to not It is detected with network environment.Probe algorithm is also that proposed algorithm defines Criterion Attribute in concentration degree sum aggregate, wherein concentration degree is used for table The diversity level for showing the result of algorithm output, when concentration degree is lower, diversity level is higher, when concentration degree is lower, Diversity level is lower；Index is concentrated to be used to indicate the measurement foundation of concentration degree, value can be time, item contents etc., Herein with no restrictions.

Matching component 304 can be offline matching component, for using user's history user behaviors log as various probe algorithms Input, in conjunction with collaborative filtering, Factorization scheduling algorithm strategy, to obtain the prediction result of different dimensions.

Analytic unit 306 can be matching result analytic unit, for compareing operational indicator, analyze user in different dimensional The conversion performance in prediction result is spent, obtains more user informations, which is drawn by the scene under one group of difference scene As composition, each scene portrait may include the scoring of behavior concentration degree, the scoring of content concentration degree etc., and the scoring of content-preference degree can It is individually scored with each side for item contents.

Formulate component 308, can be policy development component, for formulating different Generalization bounds for different user, to Scoring on different dimensions has the user of notable difference, can be according to scoring selection algorithm strategy and recommendation results, to difference Diversity of values unconspicuous user in dimension reinforces the function of detecting to preference in Generalization bounds.

Executive module 310 can be tactful application component, for the concentration degree according to strategy, preference scoring selection needle To the recommendation results and algorithm of user.In actual operation, the proposed algorithm that can be applicable according to subscriber policy selection, then to institute There is user to count, selects optimal algorithm to execute calculating to all users, calculated result is obtained, finally according to calculated result knot Share family policy selection recommendation results.

The component of above-described embodiment can be using in the product, and when being integrated with said modules in product, client can be very Easily carry out fool result optimizing.The result of final optimization pass can select different Generalization bounds for different users, keep away The problem of having exempted from all users using a Generalization bounds, and having caused to the low efficiency of user's recommended, realize raising to The technical effect of the efficiency of user's recommended, so solve due to all users using a Generalization bounds and cause to When the recommended of family, the individualized experience of user it is not high the technical issues of.

The above embodiment of the present invention can be adapted for the optimization of the recommendation effect of most of primary clients, certainly for single user The dynamic recommendation optimisation strategy generated, can be provided in recommended engine service in a manner of commercialization, and client is not necessarily to write code, Simply recommendation effect can targetedly be optimized very much.

It should be noted that for the various method embodiments described above, for simple description, therefore, it is stated as a series of Combination of actions, but those skilled in the art should understand that, the present invention is not limited by the sequence of acts described because According to the present invention, some steps may be performed in other sequences or simultaneously.Secondly, those skilled in the art should also know It knows, the embodiments described in the specification are all preferred embodiments, and related actions and modules is not necessarily of the invention It is necessary.

Through the above description of the embodiments, those skilled in the art can be understood that according to above-mentioned implementation The method of example can be realized by means of software and necessary general hardware platform, naturally it is also possible to by hardware, but it is very much In the case of the former be more preferably embodiment.Based on this understanding, technical solution of the present invention is substantially in other words to existing The part that technology contributes can be embodied in the form of software products, which is stored in a storage In medium (such as ROM/RAM, magnetic disk, CD), including some instructions are used so that a terminal device (can be mobile phone, calculate Machine, server or network equipment etc.) method that executes each embodiment of the present invention.

Embodiment 2

According to embodiments of the present invention, a kind of embodiment of object recommendation method is additionally provided, it should be noted that in attached drawing Process the step of illustrating can execute in a computer system such as a set of computer executable instructions, although also, Logical order is shown in flow charts, but in some cases, can be executed with the sequence for being different from herein it is shown or The step of description.

Embodiment of the method provided by the embodiment of the present application one can be in mobile terminal, terminal or similar fortune It calculates and is executed in device.Fig. 4 be it is according to an embodiment of the present invention it is a kind of for realizing object recommendation method terminal (or move Dynamic equipment) hardware block diagram.As shown in figure 4, terminal 400 (or mobile device 400) may include one or more A (402a, 402b ... ... being used in figure, 402n is shown) (processor 402 can include but is not limited to micro- place to processor 402 Manage the processing unit of device MCU or programmable logic device FPGA etc.), memory 404 for storing data and for communicating The transmitting device 406 of function.It in addition to this, can also include: display, input/output interface (I/O interface), general serial Port bus (USB) (a port that can be used as in the port of I/O interface is included), network interface, power supply and/or phase Machine.It will appreciated by the skilled person that structure shown in Fig. 4 is only to illustrate, not to the knot of above-mentioned electronic device It is configured to limit.For example, terminal 400 may also include than shown in Fig. 4 more perhaps less component or have with Different configuration shown in Fig. 4.

It is to be noted that said one or multiple processors 402 and/or other data processing circuits lead to herein Can often " data processing circuit " be referred to as.The data processing circuit all or part of can be presented as software, hardware, firmware Or any other combination.In addition, data processing circuit for single independent processing module or all or part of can be integrated to meter In any one in other elements in calculation machine terminal 400 (or mobile device).As involved in the embodiment of the present application, The data processing circuit controls (such as the selection for the variable resistance end path connecting with interface) as a kind of processor.

Memory 404 can be used for storing the software program and module of application software, such as the object in the embodiment of the present invention Corresponding program instruction/the data storage device of recommended method, the software that processor 402 is stored in memory 404 by operation Program and module realize the object recommendation of above-mentioned application program thereby executing various function application and data processing Method.Memory 404 may include high speed random access memory, may also include nonvolatile memory, such as one or more magnetism Storage device, flash memory or other non-volatile solid state memories.In some instances, memory 404 can further comprise phase The memory remotely located for processor 402, these remote memories can pass through network connection to terminal 400. The example of above-mentioned network includes but is not limited to internet, intranet, local area network, mobile radio communication and combinations thereof.

Transmitting device 406 is used to that data to be received or sent via a network.Above-mentioned network specific example may include The wireless network that the communication providers of terminal 400 provide.In an example, transmitting device 406 includes a network Adapter (Network Interface Controller, NIC), can be connected by base station with other network equipments so as to It is communicated with internet.In an example, transmitting device 406 can be radio frequency (Radio Frequency, RF) module, It is used to wirelessly be communicated with internet.

Display can such as touch-screen type liquid crystal display (LCD), the liquid crystal display aloow user with The user interface of terminal 400 (or mobile device) interacts.

Herein it should be noted that in some optional embodiments, above-mentioned computer equipment shown in Fig. 4 (or movement is set It is standby) it may include hardware element (including circuit), software element (including the computer generation that may be stored on the computer-readable medium Code) or both hardware element and software element combination.It should be pointed out that Fig. 4 is only a reality of particular embodiment Example, and it is intended to show that the type for the component that may be present in above-mentioned computer equipment (or mobile device).

Under above-mentioned running environment, this application provides object recommendation methods as shown in Figure 5.It should be noted that should The object recommendation method of embodiment can as shown in Figure 1 embodiment system execute.

Fig. 5 is a kind of flow chart of object recommendation method according to an embodiment of the present invention.As shown in figure 5, this method includes Following steps:

Step S502 determines the Generalization bounds for carrying out object recommendation to user according to the characteristic factor of user preference.

In the technical solution that above-mentioned steps S502 of the present invention is provided, user preference can consider object (example for user Such as, commodity and service) when made rationality have tendentious selection.The characteristic factor of user preference is for true The factor for determining user preference may include user for being recommended the preference factor of object, for example, the real-time influence recommended is used For the user preference of recommended object, it is inclined for the user for being recommended object that the accuracy of the content of recommendation influences user at family Good, for diversity influence user for the user preference of recommended object, evaluating quality influences user for the use of recommended object Family preference can be determined, for example, in concentration degree scoring, content set by the scoring of behavior concentration degree, the scoring of content concentration degree Degree scoring is higher, then it is better to evaluate, and concentration degree scoring, the scoring of content concentration degree are lower, then evaluates worse；It is deposited in recommended object When being associated with mating recommendation, user also will affect for the user preference of recommended object.The Generalization bounds of the embodiment are used for To user's recommended, by Generalization bounds object recommended to the user, that is, recommended object, different users are corresponding not Same Generalization bounds.According to the characteristic factor of user preference, determines the Generalization bounds for carrying out object recommendation to user, avoid User used is all suitable for the same Generalization bounds.

Step S504, according to the Generalization bounds determined to user's recommended.

In the technical solution that above-mentioned steps S504 of the present invention is provided, in the characteristic factor according to user preference, determines and use After the Generalization bounds for carrying out object recommendation to user, according to the Generalization bounds determined to user's recommended, which is It can satisfy the object of user demand.Optionally, had according to the corresponding probe algorithm of each of Generalization bounds characteristic factor Body selects the recommendation results and algorithm that are directed to user, for example, being directed to and using for the concentration degree of Generalization bounds, preference scoring selection The recommendation results and algorithm at family.The proposed algorithm for being applicable in user can also be selected according to Generalization bounds, then all users are carried out Statistics selects optimal algorithm to execute calculating to the Generalization bounds of all users, obtains calculated result, combines further according to calculated result To the Generalization bounds of user's progress object recommendation to user's recommended, so as to while great simplification optimization process Guarantee certain effect of optimization, can satisfy the optimization demand of naive user, enhances this crowd of user to the confidence of recommended engine, Thereby reduce the turnover rate of client.

As an optional implementation manner, characteristic factor of the step S504 according to user preference is determined for user The Generalization bounds for carrying out object recommendation include: the historical behavior based on user, determine the one or more features of user preference because Element；According to the one or more features factor of the user preference determined, the recommendation plan for carrying out object recommendation to user is determined Slightly.

It, can be with when determining the Generalization bounds for carrying out object recommendation to user in the characteristic factor according to user preference Historical behavior based on user determines the one or more features factor of user preference.It is available when obtaining user behavior The historical behavior of user can be with for example, obtain the historical log information of user, history access information, historical operation behavior etc. Parse the historical behavior of determining user to the uniform resource locator of user's access, by the answer result of user's answer come The historical behavior for obtaining user, can also collect the history of user by user some scorings, recommendation information in net etc. Behavior etc., herein with no restrictions.After obtaining user behavior, determine that the one or more of user preference is special according to user behavior Sign factor, this feature factor can be factor for influencing the feature of user preference, and then determine for carrying out pair to user As the Generalization bounds of recommendation.

As an optional implementation manner, the one or more features factor according to the user preference determined is determined and is used In the power that the Generalization bounds for carrying out object recommendation to user include: the corresponding recommendation mechanisms of each determining characteristic factor respectively Weight；According to the corresponding recommendation mechanisms of each characteristic factor and the weight of each recommendation mechanisms, determine for being carried out to user The Generalization bounds of object recommendation.

In the one or more features factor according to the user preference determined, determine for carrying out object recommendation to user When Generalization bounds, the weight of the corresponding recommendation mechanisms of each characteristic factor is determined respectively；It is corresponding according to each characteristic factor Recommendation mechanisms and each recommendation mechanisms weight, determine for user carry out object recommendation Generalization bounds.

There are recommendation mechanisms based on the formulation of each characteristic factor, each characteristic factor can formulate different recommendation machines System, that is, can have one group of recommendation mechanisms, which can be proposed algorithm, that is, can have one group of algorithm. Each recommendation mechanisms are used to pay close attention to some side of user, for example, to each side of item contents.Different characteristic factors With different recommendation mechanisms.According to determine user preference one or more features factor, determine for user into When the Generalization bounds of row object recommendation, the weight of the corresponding recommendation mechanisms of each characteristic factor is determined respectively, that is, being each The corresponding proposed algorithm of a characteristic factor defines weight, then according to the corresponding recommendation mechanisms of each characteristic factor and each pushes away The weight of mechanism is recommended, determines the Generalization bounds for carrying out object recommendation to user.Different user has different Generalization bounds, Selection algorithm strategy and recommendation results can be carried out according to scoring, for not to the apparent user of diversity of values on different dimensions With the unconspicuous user of diversity of values in dimension, have the function of reinforcing preference detection in Generalization bounds.

As an optional implementation manner, the characteristic factor of user preference includes following one: user is for recommended The preference factor of the real-time of object；Preference factor of the user for the accuracy of the content of recommended object；User for into The preference factor for the recommendation period that row is recommended；User is for the preference factor for the network environment recommended；User is for being pushed away Recommend the multifarious preference factor of object；Preference factor of the user for recommended subject evaluation quality；User is for recommended pair As there is the preference factor for being associated with mating recommendation.

The characteristic factor of user preference is the factor for determining user preference, may include user for being recommended object Preference factor.Above-mentioned user for being recommended the preference factor of the real-time of object, the real-time of recommendation influence user for The preference of recommended object；Above-mentioned user for be recommended object content accuracy preference factor, the content of recommendation Accuracy influences user for the preference of recommended object；User carries out the preference factor for the recommendation period recommended Recommendation recommends the period to influence the preference of user, can be detected according to different periods；Above-mentioned user is for being recommended The preference factor of network environment, the preference of the network environment influence user recommended can be carried out according to different network environments Detection；Above-mentioned user influences user for being recommended the inclined of object for being recommended the multifarious preference factor of object, diversity It is good, diversity level can be indicated by concentration degree, wherein concentration degree is lower, and diversity level is higher, by concentrating index table Show that the measurement foundation of concentration degree, value can be time, item contents etc.；Preference of the user for recommended subject evaluation quality Factor, evaluating quality influences user for the preference of recommended object, and having evaluated bad determinations can be commented by behavior concentration degree Point, content concentration degree scoring be determined, for example, concentration degree scoring, content concentration degree scoring it is higher, then it is better to evaluate, concentrate Degree scoring, the scoring of content concentration degree are lower, then evaluate worse；There is the mating recommendation of association for being recommended object in above-mentioned user Preference factor, when recommended object exists and is associated with mating recommendation, preference of the user for recommended object.

Optionally, which can determine the corresponding preference dimension of above-mentioned preference factor by default probe algorithm, In, the characteristic factor of user preference is divided into different dimensions by a variety of preference factors, and every kind of preference factor corresponds to a kind of dimension. For example, user for being recommended the preference factor of the real-time of object, can correspond to the preference dimension of real-time, user for The preference factor of the accuracy of the content of recommended object, can correspond to preference dimension of accuracy etc., herein without limitation.

As an optional implementation manner, the weight for determining the corresponding recommendation mechanisms of each characteristic factor respectively it Before, further includes: in the following manner, the recommendation mechanisms formulated based on each characteristic factor are obtained respectively: obtaining be based on respectively The first recommendation mechanisms or the second recommendation mechanisms that each characteristic factor is formulated, wherein the first recommendation mechanisms are characterized factor Independence between other feature factor is higher than the recommendation mechanisms of predetermined threshold, the second recommendation mechanisms be characterized factor with it is other Independence between characteristic factor is lower than the recommendation mechanisms of predetermined threshold, and predetermined threshold is used for characteristic feature factor and other feature Factor is associated.

The embodiment obtains the first recommendation mechanisms or the second recommendation mechanisms formulated based on each characteristic factor respectively, First recommendation mechanisms are characterized the recommendation mechanisms that the independence between factor and other feature factor is higher than predetermined threshold, for example, First recommendation mechanisms are characterized factor and the mutual independent or approximately independent recommendation mechanisms of other feature factor, in this way to user Recommended it is high-efficient；Second recommendation mechanisms are characterized the independence between factor and other feature factor lower than predetermined threshold Recommendation mechanisms, this kind of mechanism is general algorithm, for example, the second recommendation mechanisms are characterized between factor and other feature factor Correlation with higher, the in this way low efficiency to user's recommended, wherein predetermined threshold is for characteristic feature factor and its Its characteristic factor is associated.

As an optional implementation manner, based on the historical behavior of user, determine that the one or more of user preference is special Sign factor includes: to be predicted using the preference pattern constructed in advance the historical behavior of user, obtains including user preference One or more features factor prediction result, wherein preference training pattern according to user's predetermined quantity historical behavior with And corresponding characteristic factor training obtains.

It is also predetermined to user when determining the one or more features factor of user preference in the historical behavior based on user The historical behavior of quantity and corresponding characteristic factor are trained, and obtain preference training pattern.The preference training pattern includes The algorithm policy write in advance, that is, probe algorithm, can predict the historical behavior of user, obtain including useful The prediction result of the one or more features factor of family preference.Optionally, preference training pattern is in detection one or more features When the corresponding preference dimension of factor, distinguish environmental context, for example, detected according to different periods, according to heterogeneous networks into Row detection.It can be inputted the log of the historical behavior of predetermined quantity as various algorithms, in conjunction with collaborative filtering, Factorization etc. Algorithm policy, to obtain the prediction result of different dimensions.

As an optional implementation manner, the weight packet of the corresponding recommendation mechanisms of each characteristic factor is determined respectively It includes: according to user information entrained in prediction result, determining the weight of the corresponding recommendation mechanisms of each characteristic factor respectively, Wherein, user information is for characterizing user to the preference size of each characteristic factor.

The prediction result of the embodiment can carry user information.In the history using the preference pattern constructed in advance to user Behavior is predicted, after obtaining prediction result, is verified, is analyzed to prediction result.Determine respectively each feature because When the weight of the corresponding recommendation mechanisms of element, it is used to characterize user to each characteristic factor entrained by prediction result The user information of preference size determines the weight of the corresponding recommendation mechanisms of each characteristic factor respectively.Optionally, above-mentioned The preference of characteristic factor is bigger, and the weight of recommendation mechanisms corresponding with this feature factor is bigger, the preference of characteristic factor Degree is just smaller, and this feature factor with the weight of corresponding recommendation mechanisms with regard to smaller, the preference of characteristic factor is bigger. Conversion performance of the user in the prediction result of different characteristic factor can be analyzed, to obtain more according to operational indicator User information.User information can be drawn a portrait by the scene under one group of difference scene and be formed, and each scene portrait may include behavior Concentration degree scoring, the scoring of content concentration degree, the scoring of content-preference degree etc., wherein the scoring of content-preference degree is for recommended pair The individual scoring of the progress of the content of elephant, herein with no restrictions.

It as an optional implementation manner, include: that foundation is each to user's recommended according to the Generalization bounds determined The weight of recommendation mechanisms determines the size of the display area of the corresponding recommendation results of each recommendation mechanisms；According to each recommendation machine The size for making the display area of corresponding recommendation results, recommendation results multi-section display corresponding to each recommendation mechanisms.

After determining the Generalization bounds for carrying out object recommendation to user, pushed away according to the Generalization bounds determined to user Recommend object.Optionally, the corresponding recommendation results of each recommendation mechanisms have corresponding display area, and different recommendation results are corresponding Different size of display area.Power when according to the Generalization bounds determined to user's recommended, according to each recommendation mechanisms Weight, determines the size of the display area of the corresponding recommendation results of each recommendation mechanisms, according to the corresponding recommendation of each recommendation mechanisms As a result the size of display area.It, can after the size according to the display area of the corresponding recommendation results of each recommendation mechanisms To carry out multi-section display to the corresponding recommendation results of each recommendation mechanisms.

The embodiment passes through the characteristic factor according to user preference, determines the recommendation plan for carrying out object recommendation to user Slightly；According to the Generalization bounds determined to user's recommended.It is personalized due to that can be determined according to the characteristic factor of user preference Generalization bounds come to user's recommended, and client is without writing code, to realize the effect improved to user's recommended The technical effect of rate, and then solve when being caused using a Generalization bounds to user's recommended due to all users, it uses The individualized experience at family it is not high the technical issues of.

Under above-mentioned running environment, present invention also provides object recommendation methods as shown in FIG. 6.It should be noted that The object recommendation method of the embodiment can as shown in Figure 2 embodiment system execute.

Fig. 6 is the flow chart of another object recommendation method according to an embodiment of the present invention.As shown in fig. 6, this method packet Include following steps:

Step S602 receives the input information for retrieval.

In the technical solution that above-mentioned steps S602 of the present invention is provided, the input information for retrieval, input letter are received Breath is the information for retrieval, can be using the historical behavior log of user as the input information of various recommendation mechanisms, the input Information is corresponding with the user information of user.

Step S604 determines the user information of the corresponding user of input information.

In the technical solution that above-mentioned steps S604 of the present invention is provided, after receiving the input information for retrieval, really Surely the user information of the corresponding user of input information, the user information can be used for characterizing user to the inclined of each characteristic factor Good degree size can be drawn a portrait by the scene under one group of difference scene and be formed, and each scene portrait may include behavior concentration degree Scoring, the scoring of content concentration degree, the scoring of content-preference degree etc., wherein the scoring of content-preference degree can be for for recommended object Content progress individual scoring, herein with no restrictions.

Step S606 obtains Generalization bounds corresponding with user information in advance, wherein Generalization bounds according to user information For the characteristic factor according to user preference, object recommendation is carried out to user.

In the technical solution that above-mentioned steps S606 of the present invention is provided, in the user's letter for determining the corresponding user of input information After breath, according to user information, Generalization bounds corresponding with user information in advance are obtained.Different user is corresponding to be used for user The Generalization bounds for carrying out object recommendation are different, which is used for the characteristic factor according to user preference, carry out pair to user As recommending.

Step S608, according to the Generalization bounds of acquisition to user's recommended.

In the technical solution that above-mentioned steps S608 of the present invention is provided, recommendation plan corresponding with user information in advance is being obtained After slightly, according to the Generalization bounds of acquisition to user's recommended, which can be article, herein with no restrictions.

As an optional implementation manner, according to user information, recommendation plan corresponding with user information in advance is obtained Before slightly, further includes: in the following manner, generate Generalization bounds: the historical behavior based on user determines one of user preference Or multiple characteristic factors；The weight of the corresponding recommendation mechanisms of each characteristic factor is determined respectively；According to each characteristic factor The weight of corresponding recommendation mechanisms and each recommendation mechanisms determines the Generalization bounds for carrying out object recommendation to user.

When obtaining Generalization bounds corresponding with user information in advance, the historical behavior of available user is used for example, obtaining The historical log information at family, history access information, historical operation behavior etc., the uniform resource locator that user can also be accessed The historical behavior for parse determining user, the historical behavior of user is obtained by the answer result of user's answer, can also be with The historical behavior etc. of user is collected by some scorings, the recommendation information of user in net etc., herein with no restrictions.

After the historical behavior for obtaining user, the one or more features factor of user preference is determined.User preference pair Answer one or more features factor.Optionally, the characteristic factor of the user preference may include user for being recommended object Preference factor may include a variety of preference factors, herein with no restrictions.

It is also predetermined to user when determining the one or more features factor of user preference in the historical behavior based on user The above-mentioned historical behavior of quantity and corresponding features described above factor are trained, and obtain preference training pattern.Preference training Model includes the algorithm policy write in advance, can predict the historical behavior of user, obtain including user preference One or more features factor prediction result.

It as an optional implementation manner, include: that foundation is each to user's recommended according to the Generalization bounds of acquisition The weight of recommendation mechanisms determines the size of the display area of the corresponding recommendation results of each recommendation mechanisms；According to each recommendation machine The size for making the display area of corresponding recommendation results, recommendation results multi-section display corresponding to each recommendation mechanisms.

The embodiment is after determining the Generalization bounds for carrying out object recommendation to user, according to the Generalization bounds determined To user's recommended.Optionally, the corresponding recommendation results of each recommendation mechanisms have corresponding display area, different recommendations As a result different size of display area is corresponded to.When according to the Generalization bounds determined to user's recommended, according to each recommendation The weight of mechanism determines the size of the display area of the corresponding recommendation results of each recommendation mechanisms, according to each recommendation mechanisms pair The size of the display area for the recommendation results answered can carry out multi-section display to the corresponding recommendation results of each recommendation mechanisms.

The recommendation results of different user and algorithm difference.Optionally, according to determine Generalization bounds to user recommend pair As when, according to the weight of the corresponding recommendation mechanisms of each of Generalization bounds characteristic factor specific choice be directed to user recommendation As a result and algorithm, for example, being directed to the recommendation results and algorithm of user for the concentration degree of Generalization bounds, preference scoring selection. In actual operation, it can be applicable in the proposed algorithm of user according to subscriber policy selection, then all users is counted, selection is most Excellent algorithm executes calculating to the Generalization bounds of all users, obtains calculated result, combines further according to calculated result to user and carries out The Generalization bounds of object recommendation are to user's recommended, so as to guarantee centainly while great simplified optimization process Effect of optimization can satisfy the optimization demand of naive user, enhances this crowd of user to the confidence of recommended engine, thereby reduces The turnover rate of client.

The embodiment is by receiving the input information for retrieval；Determine the user information of the corresponding user of input information； According to user information, Generalization bounds corresponding with user information in advance are obtained, wherein Generalization bounds are used for according to user preference Characteristic factor carries out object recommendation to user；According to the Generalization bounds of acquisition to user's recommended, due to can be according to user The characteristic factor of preference determines Personalized Recommendation Strategy to user's recommended, and client is without writing code, to realize Improve the technical effect of the efficiency to user's recommended, so solve due to all users using Generalization bounds and When causing to user's recommended, the individualized experience of user it is not high the technical issues of.

It is illustrated below with reference to object recommendation method of the preferred embodiment to the embodiment of the present invention.

Fig. 7 is the flow chart of another object recommendation method according to an embodiment of the present invention.As shown in fig. 7, this method can With the following steps are included:

Step S702 classifies to proposed algorithm according to probe algorithm.

The probe algorithm of user preference is detected, and is classified according to the dimension of user preference to proposed algorithm.The example It can be classified according to probe algorithm to proposed algorithm.Wherein, probe algorithm is the algorithm write in advance.Probe algorithm inspection Survey following preference dimension, user for real-time item preference, user for item contents preference, herein with no restrictions.It visits Method of determining and calculating is also in detection preference dimension time zone time-sharing environment context, for example, detecting according to different periods, according to different network environments Detection.Probe algorithm is also that proposed algorithm defines Criterion Attribute in concentration degree sum aggregate, wherein concentration degree is for indicating algorithm output Result diversity level, when concentration degree is lower, diversity level is higher, when concentration degree is lower, diversity level It is lower；It concentrates index to be used to indicate that the measurement foundation of concentration degree, value can be time, item contents etc., does not limit herein System.

Step S704, it is offline to match, obtain the matching result of different dimensions.

After being classified according to probe algorithm to proposed algorithm, calculated user's history user behaviors log as various detections The input of method, in conjunction with collaborative filtering, Factorization scheduling algorithm strategy, so that the matching result of different dimensions is obtained, that is, obtaining Prediction result.

Step S706, analyzes matching result, obtains user information.

It is being matched offline, after obtaining the matching result of different dimensions, is compareing operational indicator, analyzing user in difference Conversion performance in dimension prediction result, obtains more user informations, the user information is by the scene under one group of difference scene It draws a portrait and forms, each scene portrait may include the scoring of behavior concentration degree, the scoring of content concentration degree etc., and content-preference degree scores, It can individually be scored for each side of item contents.

Step S708 formulates Generalization bounds.

It is analyzed to matching result, after obtaining user information, formulates different Generalization bounds for different user, it is right Scoring on different dimensions has the user of notable difference, can be according to scoring selection algorithm strategy and recommendation results, to not Reinforce the function of detecting to preference with the unconspicuous user of diversity of values in dimension, in Generalization bounds.

Step S710 executes Generalization bounds.

After formulating Generalization bounds, measured according to the concentration degree of strategy, preference scoring selection for the recommendation of user And algorithm.In actual operation, the proposed algorithm that can be applicable according to subscriber policy selection, then counts all users, Selection optimal algorithm executes calculating to all users, obtains calculated result, is finally selected according to calculated result combination subscriber policy Recommendation results.

The object recommendation method of above-described embodiment can make client that can very easily carry out fool result excellent Change.The result of final optimization pass can select different Generalization bounds for different users, avoid all users using a recommendation Strategy, and the problem of cause to the low efficiency of user's recommended, realize the technology improved to the efficiency of user's recommended Effect, and then solve when being caused using a Generalization bounds to user's recommended due to all users, the individual character of user Change and experiences the technical issues of not high.

The above embodiment of the present invention can be adapted for the optimization of the recommendation effect of most of primary clients, certainly for single user The dynamic recommendation optimisation strategy generated, can be provided in recommended engine service in a manner of commercialization, and client is not necessarily to write code, Simply recommendation effect can targetedly be optimized very much.

It should be noted that for the various method embodiments described above, for simple description, therefore, it is stated as a series of Combination of actions, but those skilled in the art should understand that, the present invention is not limited by the sequence of acts described because According to the present invention, some steps may be performed in other sequences or simultaneously.Secondly, those skilled in the art should also know It knows, the embodiments described in the specification are all preferred embodiments, and related actions and modules is not necessarily of the invention It is necessary.

Through the above description of the embodiments, those skilled in the art can be understood that according to above-mentioned implementation The method of example can be realized by means of software and necessary general hardware platform, naturally it is also possible to by hardware, but it is very much In the case of the former be more preferably embodiment.Based on this understanding, technical solution of the present invention is substantially in other words to existing The part that technology contributes can be embodied in the form of software products, which is stored in a storage In medium (such as ROM/RAM, magnetic disk, CD), including some instructions are used so that a terminal device (can be mobile phone, calculate Machine, server or network equipment etc.) execute method described in each embodiment of the present invention.

Embodiment 3

According to embodiments of the present invention, it additionally provides a kind of for implementing the object of above-mentioned object recommendation method shown in fig. 5 Recommendation apparatus.Fig. 8 is a kind of schematic diagram of object recommendation device according to an embodiment of the present invention.As shown in figure 8, the device 800 It include: the first determining module 802 and the first recommending module 804.

First determining module 802 is determined for the characteristic factor according to user preference for carrying out object recommendation to user Generalization bounds.

First recommending module 804, for according to the Generalization bounds determined to user's recommended.

Optionally, the first determining module 802 includes: the first determination unit 806, for the historical behavior based on user, really Determine the one or more features factor of user preference；Second determination unit 808, for according to one of user preference determined or Multiple characteristic factors determine the Generalization bounds for carrying out object recommendation to user.

Optionally, the second determination unit 808 include: first determine subelement 810, for determine respectively each feature because The weight of the corresponding recommendation mechanisms of element；Second determines subelement 812, for according to the corresponding recommendation mechanisms of each characteristic factor And the weight of each recommendation mechanisms, determine the Generalization bounds for carrying out object recommendation to user.

Optionally, the characteristic factor of user preference includes following one: user for be recommended object real-time it is inclined Good factor；Preference factor of the user for the accuracy of the content of recommended object；User is for recommendation period for being recommended Preference factor；User is for the preference factor for the network environment recommended；User is multifarious for recommended object partially Good factor；Preference factor of the user for recommended subject evaluation quality；User, which has recommended object, is associated with mating push away The preference factor recommended.

Optionally, the second determination unit 808 further include: obtain subelement 814, for determine respectively each feature because Before the weight of the corresponding recommendation mechanisms of element, in the following manner, the recommendation formulated based on each characteristic factor is obtained respectively Mechanism: the first recommendation mechanisms or the second recommendation mechanisms formulated based on each characteristic factor are obtained respectively, wherein first pushes away The mechanism of recommending is characterized the recommendation mechanisms that the independence between factor and other feature factor is higher than predetermined threshold, the second recommendation mechanisms The independence being characterized between factor and other feature factor is lower than the recommendation mechanisms of predetermined threshold, and predetermined threshold is for characterizing spy Sign factor is associated with other feature factor.

Optionally, the first determination unit 806 includes: prediction subelement 816, for using the preference pattern pair constructed in advance The historical behavior of user predicts, obtain include the one or more features factor of user preference prediction result, wherein Preference training pattern is obtained according to the historical behavior of user's predetermined quantity and the training of corresponding characteristic factor.

Optionally, the second determination unit 808 is also used to be determined respectively according to user information entrained in prediction result The weight of the corresponding recommendation mechanisms of each characteristic factor, wherein user information is for characterizing user to each characteristic factor Preference size.

Optionally, the first recommending module 804 includes: third determination unit 818, for the power according to each recommendation mechanisms Weight, determines the size of the display area of the corresponding recommendation results of each recommendation mechanisms；First display unit 820, for according to every The size of the display area of the corresponding recommendation results of a recommendation mechanisms, it is aobvious to the corresponding recommendation results subregion of each recommendation mechanisms Show.

Herein it should be noted that above-mentioned first determining module 802 and the first recommending module 804 correspond in embodiment 2 Step S502 and step S504, two modules are identical as example and application scenarios that corresponding step is realized, but are not limited to State 2 disclosure of that of embodiment.It should be noted that above-mentioned module may operate in embodiment 2 as a part of device In the terminal 400 of offer.

The embodiment is determined by characteristic factor of first determining module 802 according to user preference for carrying out to user The Generalization bounds of object recommendation, by the first recommending module 804 according to determining Generalization bounds to user's recommended.Due to can Personalized Recommendation Strategy is determined to user's recommended with the characteristic factor according to user preference, and client is without writing generation Code to realize the technical effect improved to the efficiency of user's recommended, and then is solved since all users are using one A Generalization bounds and when causing to user's recommended, the individualized experience of user it is not high the technical issues of.

According to embodiments of the present invention, it additionally provides a kind of for implementing the object of above-mentioned object recommendation method shown in fig. 6 Recommendation apparatus.Fig. 9 is the schematic diagram of another object recommendation device according to an embodiment of the present invention.As shown in figure 9, the device 900 include: receiving module 902, the second determining module 904, acquisition module 906 and the second recommending module 908.

Receiving module 902, for receiving the input information for being used for retrieval.

Second determining module 904, for determining the user information of the corresponding user of input information.

Module 906 is obtained, for obtaining Generalization bounds corresponding with user information in advance, wherein push away according to user information Strategy is recommended for the characteristic factor according to user preference, carries out object recommendation to user.

Second recommending module 908, for according to the Generalization bounds of acquisition to user's recommended.

Optionally, which can also include: generation module 910, for according to user information, obtaining in advance and user Before the corresponding Generalization bounds of information, in the following manner, generate Generalization bounds: the historical behavior based on user determines user The one or more features factor of preference；The weight of the corresponding recommendation mechanisms of each characteristic factor is determined respectively；According to each The weight of a corresponding recommendation mechanisms of characteristic factor and each recommendation mechanisms is determined for carrying out pushing away for object recommendation to user Recommend strategy.

Optionally, the second recommending module 908 includes: the 4th determination unit 912, for the power according to each recommendation mechanisms Weight, determines the size of the display area of the corresponding recommendation results of each recommendation mechanisms；Second display unit 914, for according to every The size of the display area of the corresponding recommendation results of a recommendation mechanisms, it is aobvious to the corresponding recommendation results subregion of each recommendation mechanisms Show.

The embodiment receives the input information for retrieval by receiving module 902；It is determined by the second determining module 904 Input the user information of the corresponding user of information；By obtaining module 906 according to user information, obtain in advance with user information pair The Generalization bounds answered, wherein Generalization bounds are used for the characteristic factor according to user preference, carry out object recommendation to user；Pass through Second recommending module 908 is according to the Generalization bounds of acquisition to user's recommended.Due to can according to user preference feature because Element determines Personalized Recommendation Strategy to user's recommended, and client is without writing code, to realize raising to user The technical effect of the efficiency of recommended, so solve due to all users using a Generalization bounds and cause to push away to user When recommending object, the individualized experience of user it is not high the technical issues of.

Herein it should be noted that above-mentioned receiving module 902, the second determining module 904, acquisition module 906 and second push away Module 908 is recommended corresponding to the step S602 and step S608 in embodiment 2, example that four modules and corresponding step are realized and Application scenarios are identical, but are not limited to the above embodiments 2 disclosure of that.It should be noted that above-mentioned module is as device A part may operate in the terminal 400 of the offer of embodiment 2.

Embodiment 4

The embodiment of the present invention can provide a kind of terminal, which can be in terminal group Any one computer terminal.Optionally, in the present embodiment, above-mentioned terminal also could alternatively be mobile whole The terminal devices such as end.

Optionally, in the present embodiment, above-mentioned terminal can be located in multiple network equipments of computer network At least one network equipment.

Optionally, Figure 10 is a kind of structural block diagram of terminal according to an embodiment of the present invention.As shown in Figure 10, should Terminal A may include: one or more (one is only shown in figure) processors 1002, memory 1004 and transmission Device 1006.

Wherein, memory can be used for storing software program and module, as the Security Object in the embodiment of the present invention is recommended Corresponding program instruction/the module of method and apparatus, the software program and module that processor is stored in memory by operation, Thereby executing various function application and data processing, that is, realize above-mentioned object recommendation method.Memory may include high speed with Machine memory, can also include nonvolatile memory, such as one or more magnetic storage device, flash memory or other are non- Volatile solid-state.In some instances, memory can further comprise the memory remotely located relative to processor, These remote memories can pass through network connection to terminal A.The example of above-mentioned network include but is not limited to internet, Intranet, local area network, mobile radio communication and combinations thereof.

In the present embodiment, above-mentioned terminal can be with following steps in the object recommendation method of executing application Program code: according to the characteristic factor of user preference, the Generalization bounds for carrying out object recommendation to user are determined；According to determining Generalization bounds to user's recommended.

Processor can call the information and application program of memory storage by transmitting device, to execute following step: According to the characteristic factor of user preference, the Generalization bounds for carrying out object recommendation to user are determined；According to the recommendation plan determined Slightly to user's recommended.

Optionally, the program code of following steps can also be performed in above-mentioned processor: the historical behavior based on user, determines The one or more features factor of user preference；According to the one or more features factor of the user preference determined, determination is used for The Generalization bounds of object recommendation are carried out to user.

Optionally, the program code of following steps can also be performed in above-mentioned processor: determining each characteristic factor respectively The weight of corresponding recommendation mechanisms；According to the corresponding recommendation mechanisms of each characteristic factor and the weight of each recommendation mechanisms, Determine the Generalization bounds for carrying out object recommendation to user.

Optionally, the program code of following steps can also be performed in above-mentioned processor: determine respectively each feature because Before the weight of the corresponding recommendation mechanisms of element, in the following manner, the recommendation formulated based on each characteristic factor is obtained respectively Mechanism: the first recommendation mechanisms or the second recommendation mechanisms formulated based on each characteristic factor are obtained respectively, wherein first pushes away The mechanism of recommending is characterized the recommendation mechanisms that the independence between factor and other feature factor is higher than predetermined threshold, the second recommendation mechanisms The independence being characterized between factor and other feature factor is lower than the recommendation mechanisms of predetermined threshold, and predetermined threshold is for characterizing spy Sign factor is associated with other feature factor.

Optionally, the program code of following steps can also be performed in above-mentioned processor: using the preference pattern constructed in advance The historical behavior of user is predicted, obtain include the one or more features factor of user preference prediction result, In, preference training pattern is obtained according to the historical behavior of user's predetermined quantity and the training of corresponding characteristic factor.

Optionally, the program code of following steps can also be performed in above-mentioned processor: entrained by prediction result User information determines the weight of the corresponding recommendation mechanisms of each characteristic factor respectively, wherein user information is for characterizing user To the preference size of each characteristic factor.

Optionally, the program code of following steps can also be performed in above-mentioned processor: according to the weight of each recommendation mechanisms, Determine the size of the display area of the corresponding recommendation results of each recommendation mechanisms；According to the corresponding recommendation results of each recommendation mechanisms Display area size, recommendation results multi-section display corresponding to each recommendation mechanisms.

In the present embodiment, following steps in the object recommendation method of application program can also be performed in above-mentioned terminal Program code: receive for retrieval input information；Determine the user information of the corresponding user of input information；Believed according to user Breath obtains Generalization bounds corresponding with user information in advance, wherein and Generalization bounds are used for the characteristic factor according to user preference, Object recommendation is carried out to user；According to the Generalization bounds of acquisition to user's recommended.

Processor can call the information and application program of memory storage by transmitting device, to execute following step: According to user information, before obtaining Generalization bounds corresponding with user information in advance, in the following manner, Generalization bounds are generated: Historical behavior based on user determines the one or more features factor of user preference；Each characteristic factor pair is determined respectively The weight for the recommendation mechanisms answered；According to the corresponding recommendation mechanisms of each characteristic factor and the weight of each recommendation mechanisms, really The fixed Generalization bounds for user's progress object recommendation.

Optionally, the program code of following steps can also be performed in above-mentioned processor: according to the weight of each recommendation mechanisms, Determine the size of the display area of the corresponding recommendation results of each recommendation mechanisms；According to the corresponding recommendation results of each recommendation mechanisms Display area size, recommendation results multi-section display corresponding to each recommendation mechanisms.

Using the embodiment of the present invention, a kind of scheme of object recommendation method is provided.The characteristic factor of foundation user preference, Determine the Generalization bounds for carrying out object recommendation to user；According to the Generalization bounds determined to user's recommended.Due to can Personalized Recommendation Strategy is determined to user's recommended with the characteristic factor according to user preference, and client is without writing generation Code to realize the technical effect improved to the efficiency of user's recommended, and then is solved since all users are using one A Generalization bounds and when causing to user's recommended, the individualized experience of user it is not high the technical issues of.

It will appreciated by the skilled person that structure shown in Fig. 10 is only to illustrate, terminal is also possible to Smart phone (such as Android phone, iOS mobile phone), tablet computer, applause computer and mobile internet device (Mobile Internet Devices, MID), the terminal devices such as PAD.Figure 10 it does not cause to limit to the structure of above-mentioned electronic device.Example Such as, terminal 400 may also include more or less component (such as network interface, display device than shown in Figure 10 Deng), or with the configuration different from shown in Figure 10.

Those of ordinary skill in the art will appreciate that all or part of the steps in the various methods of above-described embodiment is can It is completed with instructing the relevant hardware of terminal device by program, which can store in a computer readable storage medium In, storage medium may include: flash disk, read-only memory (Read-Only Memory, ROM), random access device (Random Access Memory, RAM), disk or CD etc..

Embodiment 5

The embodiments of the present invention also provide a kind of storage mediums.Optionally, in the present embodiment, above-mentioned storage medium can For saving program code performed by object recommendation method provided by above-described embodiment 2.

Optionally, in the present embodiment, above-mentioned storage medium can be located in computer network in computer terminal group In any one terminal, or in any one mobile terminal in mobile terminal group.

Optionally, in the present embodiment, storage medium is arranged to store the program code for executing following steps: according to According to the characteristic factor of user preference, the Generalization bounds for carrying out object recommendation to user are determined；According to the Generalization bounds determined To user's recommended.

Optionally, in the present embodiment, storage medium is also configured to store the program code for executing following steps: Receive the input information for retrieval；Determine the user information of the corresponding user of input information；According to user information, obtain preparatory Generalization bounds corresponding with user information, wherein Generalization bounds are used for the characteristic factor according to user preference, carry out pair to user As recommending；According to the Generalization bounds of acquisition to user's recommended.

The serial number of the above embodiments of the invention is only for description, does not represent the advantages or disadvantages of the embodiments.

In the above embodiment of the invention, it all emphasizes particularly on different fields to the description of each embodiment, does not have in some embodiment The part of detailed description, reference can be made to the related descriptions of other embodiments.

In several embodiments provided herein, it should be understood that disclosed technology contents can pass through others Mode is realized.Wherein, the apparatus embodiments described above are merely exemplary, such as the division of the unit, only A kind of logical function partition, there may be another division manner in actual implementation, for example, multiple units or components can combine or Person is desirably integrated into another system, or some features can be ignored or not executed.Another point, shown or discussed is mutual Between coupling, direct-coupling or communication connection can be through some interfaces, the INDIRECT COUPLING or communication link of unit or module It connects, can be electrical or other forms.

The unit as illustrated by the separation member may or may not be physically separated, aobvious as unit The component shown may or may not be physical unit, it can and it is in one place, or may be distributed over multiple In network unit.It can select some or all of unit therein according to the actual needs to realize the mesh of this embodiment scheme 's.

It, can also be in addition, the functional units in various embodiments of the present invention may be integrated into one processing unit It is that each unit physically exists alone, can also be integrated in one unit with two or more units.Above-mentioned integrated list Member both can take the form of hardware realization, can also realize in the form of software functional units.

If the integrated unit is realized in the form of SFU software functional unit and sells or use as independent product When, it can store in a computer readable storage medium.Based on this understanding, technical solution of the present invention is substantially The all or part of the part that contributes to existing technology or the technical solution can be in the form of software products in other words It embodies, which is stored in a storage medium, including some instructions are used so that a computer Equipment (can for personal computer, server or network equipment etc.) execute each embodiment the method for the present invention whole or Part steps.And storage medium above-mentioned includes: that USB flash disk, read-only memory (ROM, Read-Only Memory), arbitrary access are deposited Reservoir (RAM, Random Access Memory), mobile hard disk, magnetic or disk etc. be various to can store program code Medium.

The above is only a preferred embodiment of the present invention, it is noted that for the ordinary skill people of the art For member, various improvements and modifications may be made without departing from the principle of the present invention, these improvements and modifications are also answered It is considered as protection scope of the present invention.

Claims (25)

1. a kind of object recommendation method characterized by comprising

According to the characteristic factor of user preference, the Generalization bounds for carrying out object recommendation to the user are determined；

According to the Generalization bounds determined to user's recommended.

2. the method according to claim 1, wherein the characteristic factor according to the user preference, determines For including: to the Generalization bounds of user progress object recommendation

Based on the historical behavior of the user, the one or more features factor of the user preference is determined；

According to the one or more features factor of the user preference determined, determine for carrying out object recommendation to the user Generalization bounds.

3. according to the method described in claim 2, it is characterized in that, the one or more according to the determining user preference is special Sign factor determines that the Generalization bounds for carrying out object recommendation to the user include:

The weight of the corresponding recommendation mechanisms of each characteristic factor is determined respectively；

According to the corresponding recommendation mechanisms of each characteristic factor and the weight of each recommendation mechanisms, determine for described The Generalization bounds of user's progress object recommendation.

4. according to the method described in claim 3, it is characterized in that, the characteristic factor of the user preference include it is following it One:

Preference factor of the user for the real-time of recommended object；

Preference factor of the user for the accuracy of the content of recommended object；

Preference factor of the user for the recommendation period recommended；

The user is for the preference factor for the network environment recommended；

The user is for being recommended the multifarious preference factor of object；

Preference factor of the user for recommended subject evaluation quality；

There is the preference factor for being associated with mating recommendation for being recommended object in the user.

5. according to the method described in claim 3, it is characterized in that, determining the corresponding recommendation machine of each characteristic factor respectively Before the weight of system, further includes: in the following manner, obtain the recommendation mechanisms formulated based on each characteristic factor respectively:

The first recommendation mechanisms or the second recommendation mechanisms formulated based on each characteristic factor are obtained respectively, wherein described the One recommendation mechanisms be characterized the independence between factor and other feature factor be higher than predetermined threshold recommendation mechanisms, described second Recommendation mechanisms are characterized the recommendation mechanisms that the independence between factor and other feature factor is lower than the predetermined threshold, described pre- It is associated with other feature factor for characterizing the characteristic factor to determine threshold value.

6. according to the method described in claim 2, it is characterized in that, the historical behavior based on the user, determines the user The one or more features factor of preference includes:

The historical behavior of the user is predicted using the preference training pattern constructed in advance, obtains including the user The prediction result of the one or more features factor of preference, wherein the preference training pattern is according to user's predetermined quantity Historical behavior and corresponding characteristic factor training obtain.

7. according to the method described in claim 6, it is characterized in that, determining the corresponding recommendation mechanisms of each characteristic factor respectively Weight include:

According to user information entrained in the prediction result, the corresponding recommendation mechanisms of each characteristic factor are determined respectively Weight, wherein the user information is for characterizing the user to the preference size of each characteristic factor.

8. the method according to any one of claim 3 to 7, which is characterized in that according to determine the Generalization bounds to User's recommended includes:

According to the weight of each recommendation mechanisms, the size of the display area of the corresponding recommendation results of each recommendation mechanisms is determined；

The size of display area according to the corresponding recommendation results of each recommendation mechanisms, recommendation knot corresponding to each recommendation mechanisms Fruit multi-section display.

9. a kind of object recommendation method characterized by comprising

Receive the input information for retrieval；

Determine the user information of the corresponding user of the input information；

According to the user information, Generalization bounds corresponding with the user information in advance are obtained, wherein the Generalization bounds are used In the characteristic factor according to the user preference, Xiang Suoshu user carries out object recommendation；

According to the Generalization bounds of acquisition to user's recommended.

10. according to the method described in claim 9, it is characterized in that, according to the user information, obtain in advance with the use Before the corresponding Generalization bounds of family information, further includes: in the following manner, generate the Generalization bounds:

Based on the historical behavior of the user, the one or more features factor of the user preference is determined；

The weight of the corresponding recommendation mechanisms of each characteristic factor is determined respectively；

According to the corresponding recommendation mechanisms of each characteristic factor and the weight of each recommendation mechanisms, determine for described The Generalization bounds of user's progress object recommendation.

11. according to the method described in claim 10, it is characterized in that, the Generalization bounds according to acquisition are pushed away to the user Recommending object includes:

According to the weight of each recommendation mechanisms, the size of the display area of the corresponding recommendation results of each recommendation mechanisms is determined；

The size of display area according to the corresponding recommendation results of each recommendation mechanisms, recommendation knot corresponding to each recommendation mechanisms Fruit multi-section display.

12. a kind of object recommendation device characterized by comprising

First determining module is determined for the characteristic factor according to user preference for carrying out object recommendation to the user Generalization bounds；

First recommending module, for according to the Generalization bounds determined to user's recommended.

13. device according to claim 12, which is characterized in that first determining module includes:

First determination unit determines the one or more features of the user preference for the historical behavior based on the user Factor；

Second determination unit is determined for the one or more features factor according to the user preference determined for institute State the Generalization bounds that user carries out object recommendation.

14. device according to claim 13, which is characterized in that second determination unit includes:

First determines subelement, for determining the weight of the corresponding recommendation mechanisms of each characteristic factor respectively；

Second determines subelement, for according to the corresponding recommendation mechanisms of each characteristic factor and each recommendation mechanisms Weight determines the Generalization bounds for carrying out object recommendation to the user.

15. device according to claim 14, which is characterized in that the characteristic factor of the user preference includes following One of:

Preference factor of the user for the real-time of recommended object；

Preference factor of the user for the accuracy of the content of recommended object；

Preference factor of the user for the recommendation period recommended；

The user is for the preference factor for the network environment recommended；

The user is for being recommended the multifarious preference factor of object；

Preference factor of the user for recommended subject evaluation quality；

There is the preference factor for being associated with mating recommendation for being recommended object in the user.

16. device according to claim 14, which is characterized in that further include: subelement is obtained, for determining respectively often Before the weight of the corresponding recommendation mechanisms of one characteristic factor, in the following manner, obtains be based on each characteristic factor respectively The recommendation mechanisms of formulation:

The first recommendation mechanisms or the second recommendation mechanisms formulated based on each characteristic factor are obtained respectively, wherein described the One recommendation mechanisms be characterized the independence between factor and other feature factor be higher than predetermined threshold recommendation mechanisms, described second Recommendation mechanisms are characterized the recommendation mechanisms that the independence between factor and other feature factor is lower than the predetermined threshold, described pre- It is associated with other feature factor for characterizing the characteristic factor to determine threshold value.

17. device according to claim 13, which is characterized in that first determination unit includes:

It predicts subelement, for being predicted using the preference training pattern constructed in advance the historical behavior of the user, obtains To the prediction result for the one or more features factor for including the user preference, wherein the preference training pattern foundation The historical behavior of user's predetermined quantity and the training of corresponding characteristic factor obtain.

18. device according to claim 17, which is characterized in that second determination unit is also used to according to described pre- User information entrained in result is surveyed, determines the weight of the corresponding recommendation mechanisms of each characteristic factor respectively, wherein described User information is for characterizing the user to the preference size of each characteristic factor.

19. device described in any one of 4 to 18 according to claim 1, which is characterized in that first recommending module includes:

Third determination unit determines the corresponding recommendation results of each recommendation mechanisms for the weight according to each recommendation mechanisms The size of display area；

First display unit is pushed away for the size of the display area according to the corresponding recommendation results of each recommendation mechanisms to each Recommend the corresponding recommendation results multi-section display of mechanism.

20. a kind of object recommendation device characterized by comprising

Receiving module, for receiving the input information for being used for retrieval；

Second determining module, for determining the user information of the corresponding user of the input information；

Module is obtained, for obtaining Generalization bounds corresponding with the user information in advance according to the user information, wherein The Generalization bounds carry out object recommendation for the characteristic factor according to the user preference, Xiang Suoshu user；

Second recommending module, for according to the Generalization bounds of acquisition to user's recommended.

21. device according to claim 20, which is characterized in that further include: generation module, for according to the user Information in the following manner, generates the Generalization bounds before obtaining Generalization bounds corresponding with the user information in advance:

Based on the historical behavior of the user, the one or more features factor of the user preference is determined；

The weight of the corresponding recommendation mechanisms of each characteristic factor is determined respectively；

According to the corresponding recommendation mechanisms of each characteristic factor and the weight of each recommendation mechanisms, determine for described The Generalization bounds of user's progress object recommendation.

22. device according to claim 21, which is characterized in that second recommending module includes:

4th determination unit determines the corresponding recommendation results of each recommendation mechanisms for the weight according to each recommendation mechanisms The size of display area；

Second display unit is pushed away for the size of the display area according to the corresponding recommendation results of each recommendation mechanisms to each Recommend the corresponding recommendation results multi-section display of mechanism.

23. a kind of storage medium, which is characterized in that the storage medium includes the program of storage, wherein run in described program When control the storage medium where equipment perform claim require any one of 1 to 11 described in object recommendation method.

24. a kind of processor, which is characterized in that the processor is for running program, wherein right of execution when described program is run Benefit require any one of 1 to 11 described in object recommendation method.

25. a kind of system characterized by comprising

Processor；And

Memory is connected to the processor, for providing the instruction for handling following processing step for the processor:

Step 1, according to the characteristic factor of user preference, the Generalization bounds for carrying out object recommendation to the user are determined；

Step 2, according to the Generalization bounds determined to user's recommended.