Be user user_iTo all commodity classifications and the commodity class number of clicks of commodity now, U_iBe user user_iThe set of all commodity classification numberings of clicking;

B1: calculate each commodity classification cat_jThe probability P of on average being clicked_Cat(cat_j):

Because the welcome degree of each commodity classification is different; The number of times that the classification that has is unavoidably clicked is than higher, and the number of times that the classification that has is clicked is lower, clicks the influence that user click frequency is judged classification interest in order to eliminate classification; Introduced classification and clicked probability, computing method are following:

P_{cat} ({cat}_{j}) = \frac{{count}_{0} ({cat}_{j})}{\underset{k &Element; U_{0}}{Σ} {count}_{0} ({cat}_{k})}

Wherein, P_Cat(cat_j) be commodity classification cat_jThe probability of on average being clicked, count₀(cat_j) be commodity classification cat_jThe quantity of following all commodity,

Be all commodity classes total quantity of commodity now, U₀Set for all commodity classification numberings;

C1: calculate the probability P that the user clicks each commodity classification at random_Randomcat(cat_j):

Suppose that the user does not have interested commodity classification, promptly the user is at random to the click behavior of each classification, thinks that the user is evenly distribution to the click of each classification this moment.Calculate when classification is evenly distributed and do not consider the click-through count of user to classification, only consider the number of the classification of user capture, account form is following:

P_{randomcat} ({cat}_{j}) = \frac{1}{count (U_{0})}

Wherein, P_Randomcat(cat_j) be commodity classification cat_jBy the probability that the user clicks at random, count (U₀) be the number of commodity classification, wherein, U₀Set for all commodity classification numberings; For example, certain website has 3 commodity classifications, then the user to click the probability of each commodity classification at random be 1/3.

D1: judge P_Usercat(user_i, cat_j)＞P_Randomcat(cat_j)+P_Cat(cat_j) whether set up, if, then with commodity classification cat_jConfirm as user user_iInterest commodity classification.

In the formula of above step, i representes user's numbering, and j representes the numbering of commodity classification, and the value of i, j generally can be represented with natural number, U₀Represent the set of all commodity classification numberings.For example, user's 1 concurrent hit cat₁, cat₂, cat₃Three commodity classifications, according to step a1, the click probability that draws 1 pair of 3 classification of user is respectively 0.8,0.1,0.1; According to step b1, obtain under the situation of not considering number of clicks, the click probability of three classifications is respectively 0.3,0.2,0.2; According to step c1, obtain the probability that the user clicks each commodity classification at random and be 1/3; So, according to steps d 1, can obtain for cat₁, inequality is set up, and therefore can confirm that user 1 interest commodity classification is cat₁

Because the application's scheme is to carry out commercial product recommending to different user (corresponding different i values), and the P among the above-mentioned steps b1_Cat(cat_j) with step c1 in P_Randomcat(cat_j) have nothing to do with the user, therefore, can be in advance to P_Cat(cat_j) and P_Randomcat(cat_j) value store.Promptly need in each recommendation process, all not carry out the calculation procedure of b1 and c1, and only when commodity classification or commodity class commodity now change, again to P_Cat(cat_j), P_Randomcat(cat_j) numerical value upgrade.

In addition, it is understandable that the execution sequence of above-mentioned step a1, b1, c1 also can exchange, perhaps carry out side by side that these do not influence the realization of present embodiment technical scheme.

2), confirm user's interest commodity classification according to user's search behavior data:

In user behavior data, generally also comprise user's search behavior data, comprise that the user used which keyword is searched for, the access times of keyword or the like.These data also can be used for confirming user's interest commodity classification, and this scheme specifically can may further comprise the steps:

A2, from said search behavior data, extract the access times of used searching key word of user and searching key word;

B2, according to the corresponding relation of searching key word and commodity, obtain the searching times of user to commodity;

In e-commerce website, modal search behavior is to search for as keyword with trade name, can think, this keyword search of the every use of user once just is equivalent to corresponding commodity have been done once search.

Certainly, in actual use, the user possibly can't use complete trade name to search for, and for example, the complete name of commodity is " X (brand) Y model mobile phone ", and so, the user only uses X or Y to search for as keyword possibly; Perhaps, for some large-scale e-commerce websites, the user also possibly search for as keyword with businessman's title etc.To this situation, for the ease of statistics, can be according to all users' historical search behavior, the frequency of utilization that selects some in advance is higher, or more representative keyword as the foundation of adding up.The all corresponding a kind of definite commodity of each selected keyword like this, in case detect the employed searching key word of user and these keywords that presets coupling, just can think that the user has done once search to corresponding commodity.

C2, according to the corresponding relation of commodity and commodity classification, statistics of user's is to the searching times of every kind of commodity classification;

D2, choose and have at least one commodity classification of the highest searching times, choosing the interest commodity classification that the result confirms as said user.

Generally speaking, user's search behavior often can directly embody user's purchase interest, therefore, carries out commercial product recommending according to user's search behavior data, also can receive effect preferably.Certainly, in practical application, can also include the situation that the direct commodity in use category name of user is referred to as searching key word the searching times statistics of user in, so that statistics is more perfect to every kind of commodity classification.

S103 in determined interest commodity classification, chooses commodity and recommends to the user.

In this step, can recommend this interest commodity class now hot item, best buy or newly advance commodity, in the final recommendation results, can comprise the relevant information of one or more commodity, these information can represent to the user with the form of recommendation list.

More than, provide two kinds of behavioral datas to carry out the concrete scheme of commercial product recommending according to the user, in practical application, these two kinds of schemes can be distinguished independent use, also can use jointly.For example; Click behavioral data and search behavior data according to the user; Obtain two groups of recommendation results respectively, system carries out overall treatment to these two groups of recommendation results, and the mode of overall treatment can be that two groups of recommendation results are showed the user respectively; And inform every group of result's recommendation foundation, let the user select according to demands of individuals; Certainly, also can be respectively from two groups of recommendation results a picked at random part, showing the user after mixing.Present embodiment does not limit concrete overall treatment mode.

In another embodiment of the application; A kind of concrete implementation of above-mentioned steps S103 can also be: carry out commercial product recommending according to the user in the buying behavior data of website; That is: after application the foregoing description scheme is confirmed certain user's interest commodity classification; Further, judge whether this user had bought the commodity in the said interest commodity classification according to this user's buying behavior data; If then according to the information of user's purchased item, choose commodity and recommend, to realize recommendation results more accurately to the user.

Wherein, according to the information of purchased item, choose commodity several kinds of schemes below the user recommends specifically can adopt:

1) calculates said purchased item and its affiliated commodity class similarity of other commodity now, choose with the highest at least a commodity of said purchased item similarity and recommend to said user;

Wherein, the similarity sim between two kinds of commodity A and the B (A, B) can calculate according to following formula:

sim (A, B) = \frac{| U_{A} \cap U_{B} |}{| U_{A} \cup U_{B} |}

U_AUser's set of commodity A, U were bought in expression_BUser's set of commodity B was bought in expression, | U_A∩ U_B| user's total number of persons of commodity A and commodity B was bought in expression, | U_A∪ U_B| user's total number of persons of commodity A or commodity B was bought in expression.

Said method is to confirm the similarity between the commodity according to the similarity of different user buying behavior, for example, has 6 users once to buy commodity A; There are 5 users once to buy commodity B; Wherein have 2 users not only to buy commodity A but also bought commodity B, so, can obtain | U_A∩ U_B|=2, | U_A∪ U_B|=9, then commodity A and B between similarity be 2/9.

Suppose that the user once bought the commodity A under the commodity classification 1, use above-mentioned formula so, can calculate the similarity of commodity A and 1 time other commodity of classification respectively.Further, the size order that can arrange according to the similarity of commodity A and 1 time other commodity of classification is chosen one or more the highest commodity of similarity, as the commodity of recommending to the active user; Certainly, also a similarity threshold value can be set in advance,, then can recommend the active user this commodity are added in the recommendation list if the similarity with commodity A of certain commodity is not less than this threshold value.

2) whether the judges purchased item is the periodicity consumer lines; If then recommend said purchased item to said user.

Some commodity is to have relatively more fixing life cycle, for example cosmetics.If the commodity that the user has bought belong to periodically consumer lines of this type, so system can this Website login time of judges with the interval of the last time of buying these commodity whether greater than the average life cycle of these commodity; If, think that then the commodity that the user was bought last time possibly finish using, new purchasing demand has been arranged, thereby can recommend this commodity once more to this user.

It is understandable that " not being the periodicity consumer lines " is a kind of build-in attribute of commodity, simultaneously,, also should comprise the attribute of " average life cycle " for the periodicity consumer lines.

3) whether the judges purchased item has associated articles; If, the associated articles of then recommending said purchased item to said user.

There are some associated articles in some commodity, and for example, for mobile phone, its associated articles possibly comprise battery, cell-phone cover of respective model or the like, and the information of these associated articles also can be used as a kind of build-in attribute of commodity.If there is associated articles in the commodity that the user bought, system can add these associated articles in the recommendation list so, recommends the active user.

More than, provide three kinds of information to carry out the concrete scheme of commercial product recommending according to purchased item, in practical application, these three kinds of schemes can be distinguished independent use, and resulting different recommendation results are added in the recommendation list respectively.Also can carry out above-mentioned three kinds of schemes in a certain order, Fig. 3 shows the process flow diagram that a kind of order is carried out the commercial product recommending method of above-mentioned three kinds of schemes, may further comprise the steps:

S201 obtains user's behavioral data;

S202 according to user's behavioral data, confirms user's interest commodity classification;

S203 according to user's buying behavior data, judges whether this user had bought the commodity in the said interest commodity classification, if, execution in step S204 then;

Wherein, the implementation of step S201 and S202 can be similar with S101 and S102.And in S203, at first whether judges had bought the commodity in the said interest commodity classification, if then carry out follow-up step, to realize recommendation results more accurately.If not, then can be directly according to the mode of S103, for example recommend this interest commodity class now hot item, best buy or newly advance commodity etc.。

S204, whether the judges purchased item is the periodicity consumer lines, if, execution in step S205, otherwise execution in step S207;

S205, whether the interval of this Website login time of user and the last time of buying these commodity greater than the average life cycle of these commodity, if, execution in step S206, otherwise, execution in step S207;

S206 recommends said purchased item to the user;

S207, whether the judges purchased item has associated articles, if, execution in step S208 then, otherwise execution in step S209;

S208, the associated articles of recommending said purchased item to the user;

S209 calculates said purchased item and its affiliated class similarity of other commodity now, chooses with the highest at least a commodity of said purchased item similarity and recommends to the user.

In the technical scheme that present embodiment provided, confirm user's interest commodity classification after, further according to this user's buying behavior data, judge whether this user had bought the commodity in the said interest commodity classification; If then according to the information of user's purchased item, choose commodity and recommend, to realize recommendation results more accurately to the user.

Corresponding to top method embodiment, the application also provides a kind of commercial product recommending system.As shown in Figure 4, a kind of commercial product recommending system that the application embodiment provides can comprise:

Behavioraldata acquisition module 510 obtains said user's behavioral data, and said behavioral data comprises: the user is at the click behavioral data of website, or the user is in the search behavior data of website;

Interestclassification determination module 520 is used for according to said behavioral data, confirms said user's interest commodity classification;

Recommendingmodule 530 is used for choosing commodity and recommending to said user at determined interest commodity classification.

Wherein, Said behavioraldata acquisition module 510; The probability that probability that specifically can be used for according to the user click probability of said commodity classification, said commodity classification being clicked by the user at random and said commodity classification are on average clicked; Probability whether judges is clicked by the user greater than said commodity classification the click probability of said commodity classification at random and said commodity classification be by the average probability sum of clicking, if then this commodity classification is confirmed as user's interest commodity classification.

Further, the user behavior data that obtains of said behavioraldata acquisition module 510 can be the click behavioral data of user in the website; As shown in Figure 5, said interestclassification determination module 520 specifically can comprise:

The user clicks probability calculation submodule 5201, is used to calculate user user_iTo each commodity classification cat_jThe click probability P_Usercat(user_i, cat_j):

P_{usercat} ({user}_{i}, {cat}_{j}) = \frac{{count}_{i} ({cat}_{j})}{\underset{k &Element; U_{i}}{Σ} {count}_{i} ({cat}_{k})}

First confirms submodule 5202, is used to judge P_Usercat(user_i, cat_j)＞P_Randomcat(cat_j)+P_Cat(cat_j) whether set up, if, then with commodity classification cat_jConfirm as user user_iInterest commodity classification;

Wherein, P_Randomcat(cat_j) be commodity classification cat_jBy the probability that the user clicks at random, P_Cat(cat_j) be commodity classification cat_jThe probability of on average being clicked.

Wherein, the user behavior data that obtains of behavioraldata acquisition module 510 can also be the search behavior data of user in the website; As shown in Figure 6, said interestclassification determination module 520 specifically can also comprise:

Keyword extraction submodule 5203 is used for from said search behavior data, extracts the access times of used searching key word of user and searching key word;

The commercial articles searching number of times obtainssubmodule 5204, is used for the corresponding relation according to searching key word and commodity, obtains the searching times of user to commodity;

The class heading search number of times obtainssubmodule 5205, is used for the corresponding relation according to commodity and commodity classification, and statistics of user's is to the searching times of every kind of commodity classification;

Second confirmssubmodule 5206, is used to choose have at least one commodity classification of the highest searching times, with choosing the interest commodity classification that the result confirms as said user.

As shown in Figure 7, recommendingmodule 530 can comprise described in the commercial product recommending system that the application embodiment provided:

Judge submodule 5301, be used for buying behavior data, judge whether this user had bought the commodity in the said interest commodity classification according to said user;

Recommend submodule 5302, be used for having bought under the situation of said interest commodity classification commodity,, choose commodity and recommend to said user according to the information of user's purchased item the user.

Said recommendation submodule 5302 specifically can be configured to: be used to calculate said purchased item and its affiliated class similarity of other commodity now, and choose with the highest at least a commodity of said purchased item similarity and recommend to said user;

Wherein, the similarity sim of commodity A and B (A, B) calculate according to following formula:

sim (A, B) = \frac{| U_{A} \cap U_{B} |}{| U_{A} \cup U_{B} |}

U_AFor buying user's set of commodity A, U_BFor the user who bought commodity B gathers,

| U_A∩ U_B| for buying user's total number of persons of commodity A and commodity B,

| U_A∪ U_B| for buying user's total number of persons of commodity A or commodity B.

Whether in another embodiment of the application, said recommendation submodule 5302: being used for the judges purchased item is the periodicity consumer lines if can also being configured to; If then recommend said purchased item to said user.Whether the interval that wherein, can also further judge this Website login time of said user and the last time of buying these commodity is greater than the average life cycle of these commodity; If then recommend said purchased item to said user.

In another embodiment of the application, said recommendation submodule 5302 can also be configured to: be used for the judges purchased item and whether have associated articles; If, the associated articles of then recommending said purchased item to said user.

More than, three kinds of concrete configuration schemes of recommending submodule 5302 are provided, in practical application; These three kinds of schemes can be distinguished independent use; With the different recommendation results that obtains, certainly, recommend submodule 5302 also can concrete configuration carry out three kinds of pairing methods of scheme for order; Specifically can be referring to the description of method embodiment part, no longer repeat specification here.

The commercial product recommending system that the application embodiment provided can be a server apparatus that independently has the commercial product recommending function, also can be a functional entity that is arranged in other server apparatus.The behavioral data of this system user is confirmed user's interest commodity classification, and chooses corresponding commodity according to user's interest commodity classification and recommend to the user.Compare with existing commercial product recommending scheme, the application embodiment scheme is the interest to different personal users, carries out commercial product recommending respectively.Therefore can reduce the personal user invalidly browse, behavior such as click, promote the shopping impression.From the angle of shopping website, except can improving the website trading volume, can also effectively alleviate the burden of Website server, save network bandwidth resources and take.

For the convenience of describing, be divided into various modules with function when describing above the device and describe respectively.Certainly, when implementing the application, can in same or a plurality of softwares and/or hardware, realize the function of each unit.

Description through above embodiment can know, those skilled in the art can be well understood to the application and can realize by the mode that software adds essential general hardware platform.Based on such understanding; The part that the application's technical scheme contributes to prior art in essence in other words can be come out with the embodied of software product; This computer software product can be stored in the storage medium, like ROM/RAM, magnetic disc, CD etc., comprises that some instructions are with so that a computer equipment (can be a personal computer; Server, the perhaps network equipment etc.) carry out the described method of some part of each embodiment of the application or embodiment.

Each embodiment in this instructions all adopts the mode of going forward one by one to describe, and identical similar part is mutually referring to getting final product between each embodiment, and each embodiment stresses all is the difference with other embodiment.Especially, for system embodiment, because it is basically similar in appearance to method embodiment, so describe fairly simplely, relevant part gets final product referring to the part explanation of method embodiment.System embodiment described above only is schematic; Wherein said unit as the separating component explanation can or can not be physically to separate also; The parts that show as the unit can be or can not be physical locations also; Promptly can be positioned at a place, perhaps also can be distributed on a plurality of NEs.Can realize the purpose of present embodiment scheme according to the needs selection some or all of module wherein of reality.Those of ordinary skills promptly can understand and implement under the situation of not paying creative work.

The application can be used in numerous general or special purpose computingasystem environment or the configuration.For example: personal computer, server computer, handheld device or portable set, plate equipment, multicomputer system, the system based on microprocessor, set top box, programmable consumer-elcetronics devices, network PC, small-size computer, mainframe computer, comprise DCE of above any system or equipment or the like.

The application can describe in the general context of the computer executable instructions of being carried out by computing machine, for example program module.Usually, program module comprises the routine carrying out particular task or realize particular abstract, program, object, assembly, data structure or the like.Also can in DCE, put into practice the application, in these DCEs, by through communication network connected teleprocessing equipment execute the task.In DCE, program module can be arranged in this locality and the remote computer storage medium that comprises memory device.

The above only is the application's a embodiment; Should be pointed out that for those skilled in the art, under the prerequisite that does not break away from the application's principle; Can also make some improvement and retouching, these improvement and retouching also should be regarded as the application's protection domain.

Claims

1. a commercial product recommending method is characterized in that, comprising:

2. method according to claim 1 is characterized in that, said user behavior data is the click behavioral data of user in said website; Said according to said behavioral data, confirm that said user's interest commodity classification comprises:

3. method according to claim 2 is characterized in that, and is said according to said behavioral data, confirms that said user's interest commodity classification comprises:

Calculate user user_iTo each commodity classification cat_jThe click probability P_Usercat(user_i, cat_j):

P_{usercat} ({user}_{i}, {cat}_{j}) = \frac{{count}_{i} ({cat}_{j})}{\underset{k &Element; U_{i}}{Σ} {count}_{i} ({cat}_{k})}

Judge P_Usercat(user_i, cat_j)＞P_Randomcat(cat_j)+P_Cat(cat_j) whether set up, if, then with commodity classification cat_jConfirm as user user_iInterest commodity classification;

4. method according to claim 1 is characterized in that, said user behavior data is: the user is in the search behavior data of said website; Said according to said behavioral data, confirm that said user's interest commodity classification comprises:

From said search behavior data, extract the access times of used searching key word of user and searching key word;

According to the corresponding relation of searching key word and commodity, obtain the searching times of user to commodity;

According to the corresponding relation of commodity and commodity classification, statistics of user's is to the searching times of every kind of commodity classification;

Choose and have at least one commodity classification of the highest searching times, with choosing the interest commodity classification that the result confirms as said user.

5. according to each described method of claim 1 to 4, it is characterized in that said user behavior data also comprises: the user is in the buying behavior data of said website; Saidly in determined interest commodity classification, choose commodity and recommend, comprising to said user:

According to said user's buying behavior data, judge whether this user had bought the commodity in the said interest commodity classification;

If, choose commodity and recommend to said user then according to the information of user's purchased item.

6. method according to claim 5 is characterized in that, said information according to user's purchased item is chosen commodity and recommended to said user, comprising:

Calculate said and purchased item and its affiliated commodity class similarity of other commodity now, choose with the highest at least a commodity of said purchased item similarity and recommend to said user;

sim (A, B) = \frac{| U_{A} \cap U_{B} |}{| U_{A} \cup U_{B} |}

7. method according to claim 5 is characterized in that, said information according to user's purchased item is chosen commodity and recommended to said user, comprising:

Whether the judges purchased item is the periodicity consumer lines; If then recommend said purchased item to said user.

8. method according to claim 7 is characterized in that, saidly recommends said purchased item to said user, comprising:

Whether the interval of judging this Website login time of said user and the last time of buying these commodity is greater than the average life cycle of these commodity; If then recommend said purchased item to said user.

9. method according to claim 5 is characterized in that, said information according to user's purchased item is chosen commodity and recommended to said user in determined interest commodity classification, comprising:

Whether the judges purchased item has associated articles; If, the associated articles of then recommending said purchased item to said user.

10. a commercial product recommending system is characterized in that, comprising:

11. system according to claim 10 is characterized in that, said user behavior data is the click behavioral data of user in the website; Said interest classification determination module; The probability that probability that specifically is used for according to the user click probability of said commodity classification, said commodity classification being clicked by the user at random and said commodity classification are on average clicked; Probability whether judges is clicked by the user greater than said commodity classification the click probability of said commodity classification at random and said commodity classification are by the average probability sum of clicking; If then this commodity classification is confirmed as user's interest commodity classification.

12. system according to claim 11 is characterized in that, said interest classification determination module specifically comprises:

The user clicks the probability calculation submodule, is used to calculate user user_iTo each commodity classification cat_jThe click probability P_Usercat(user_i, cat_j):

P_{usercat} ({user}_{i}, {cat}_{j}) = \frac{{count}_{i} ({cat}_{j})}{\underset{k &Element; U_{i}}{Σ} {count}_{i} ({cat}_{k})}

First confirms submodule, is used to judge P_Usercat(user_i, cat_j)＞P_Randomcat(cat_j)+P_Cat(cat_j) whether set up, if, then with commodity classification cat_jConfirm as user user_iInterest commodity classification;

13. system according to claim 10 is characterized in that, said user behavior data is: the user is in the search behavior data of website; Said interest classification determination module comprises:

The keyword extraction submodule is used for from said search behavior data, extracts the access times of used searching key word of user and searching key word;

The commercial articles searching number of times obtains submodule, is used for the corresponding relation according to searching key word and commodity, obtains the searching times of user to commodity;

The class heading search number of times obtains submodule, is used for the corresponding relation according to commodity and commodity classification, and statistics of user's is to the searching times of every kind of commodity classification;

Second confirms submodule, is used to choose have at least one commodity classification of the highest searching times, with choosing the interest commodity classification that the result confirms as said user.

14. according to each described system of claim 10 to 13, it is characterized in that said user behavior data also comprises: the user is in the buying behavior data of said website; Said recommending module comprises:

Judge submodule, be used for buying behavior data, judge whether this user had bought the commodity in the said interest commodity classification according to said user;

Recommend submodule, be used for having bought under the situation of said interest commodity classification commodity,, choose commodity and recommend to said user according to the information of user's purchased item the user.