CN103093369A - Method and device for offering matched product based on correlation degree between products - Google Patents
Method and device for offering matched product based on correlation degree between products Download PDFInfo
- Publication number
- CN103093369A CN103093369A CN2011103434418A CN201110343441A CN103093369A CN 103093369 A CN103093369 A CN 103093369A CN 2011103434418 A CN2011103434418 A CN 2011103434418A CN 201110343441 A CN201110343441 A CN 201110343441A CN 103093369 A CN103093369 A CN 103093369A
- Authority
- CN
- China
- Prior art keywords
- product
- collection
- collocation
- degree
- item
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention relates to the field of electronic commerce and discloses a method and a device for offering a matched product based on correlation degree between products. The method and the device for offering the matched product based on the correlation degree between the products are used for improving accuracy of the recommended matched product when a user selects one product. Namely, when the matched product of the targeted product selected by the user is determined, an affair assembly is determined based on the setting time length so that on one hand, effectiveness of formation of affairs is guaranteed and on the other hand, relevance, between the products, contained in the affairs is guaranteed. In the process of calculating the absolute support degree of monomial set and multinomial set of the products, correlative detection is introduced to compensate statistics property shortage of the absolute support degree so that when the best matched relationship between products is determined based on the correlation degree of the products, the accuracy of the matched product determined in effectively improved, and therefore when the matched product of the targeted product is shown to the user, the accurate matched product is offered in time to the user by an electronic commerce website based on requirements of the user, and electronic commerce service quality is effectively guaranteed.
Description
Technical field
The present invention relates to e-commerce field, particularly between a kind of product-based the degree of association provide the collocation product method and device.
Background technology
Day by day universal along with E-business applications, relevant e-commerce website has also obtained unprecedented development.Under prior art, choose in the webpage of user at e-commerce website (as, browse, buy or collect) during a product, the e-business network standing-meeting has the collocation product (also claiming best collocation product) of best correlation with choosing product in its recommended website, thereby facilitate the disposable discovery of the user product relevant with purchase intention, and then minimizing user's running time, the transaction processing efficient of raising e-commerce website, the service quality of lifting e-commerce website.
Obviously, the recommendation of best collocation product has been one of technological means of e-commerce website indispensability.So, the user view how to show according to historical transaction record selects best collocation product, is that each e-commerce website needs one of technological project of at first optimizing, and this will become one of important indicator of weighing the e-commerce website service level.
At present, when choosing a product in the webpage of user at e-commerce website, for fear of blindness and loaded down with trivial details search and navigation process, the e-business network standing-meeting is based on each user's who puts down in writing in historical transaction record a shopping process, find quickly and the above-mentioned the most supporting or maximally related product of product of choosing, be prompted to the user as the best collocation product, to accelerate user's validity process of decision-making of doing shopping, experience and loyalty thereby promote the user; On the other hand, also can expand user's shopping interest, excite the shopping amount, promote the clicking rate and click conversion ratio of website.In this course, the best between each product collocation relation (being correlation degree), normally determine by calculate the number of times that each product bought simultaneously in same order: the best collocation product of any products A is bought the maximum product mix of number of times simultaneously in same order with A.
The theory of this method thinks that the product mix with best collocation relation is and is only the most frequent purchased product mix in history.Based on this theory, the method for collocation relation between existing definite product (being correlation degree) is as follows:
1) form the affairs set.Be specially: the product set that each user is bought in an order is set as affairs.
2) product with purchased mistake in the website makes up arbitrarily in twos, and calculates the absolute support of each two product mix; Wherein, the absolute support of any two product mixes is the number of transactions that comprises simultaneously these two products.
3) product with purchased mistake in the website carries out any three or three combinations, and calculates the absolute support of each three product mix, and wherein, the absolute support of any three product mixes is the number of transactions that comprises simultaneously these three products.
4) for any products A, according to step 2) acquired results, obtain comprising all two product mix set (A2) of A, and the absolute support of these set.
5) for any products A, to step 4) gained A2 screens, and deletes absolute support less than two product mixes of a setting threshold u2, obtains the selection result B2.
6) for any products A, in step 5) in gained B2, select absolute support to reach two product mixes of predetermined threshold value (for example, definitely support is maximum), form the best collocation product set C2 of two product mixes of A.
7) for any products A, according to step 3) acquired results, obtain comprising all three product mix set (A3) of A, and the absolute support of these set.
8) for any products A, to step 7) gained A3 screens, and delete absolute support and obtain less than three combinations of a setting threshold u3.Filter result B3.
9) for any products A, in step 8) in gained B3, select absolute support to reach three product mixes of predetermined threshold value (for example, definitely support is maximum), form the best collocation product set C3 of three product mixes of A.
10) for any products A, if by step 9) gained C3 is not empty, select any one three product mix in C3 as the best collocation composition of relations PA1 of A, the correlation degree that is other two products except product A of comprising in PA1 and product A is the highest, when the user chooses product A, e-commerce website will be recommended other two products to the user according to the incidence relation of three product mix signs in PA1; If by step 9) gained C3 be empty and by selecting step 6) gained C2 is for empty, selects in C2 any one two product mix as the best collocation composition of relations PA2 of A; The correlation degree that is another product except product A of comprising in PA2 and product A is the highest, when the user first in during product A, e-commerce website will be recommended another product to the user according to the incidence relation that two product mixes in PA2 characterize; If C3 and C2 are all sky, think that product A is without best matched combined product.
Yet under prior art, when e-business network stood in the incidence relation of determining between product, as affairs, this processing mode did not have desirable operability in actual applications with the product set of user's single purchase.This be because, for many businessmans and deposit and launch respectively for the comprehensive e-commerce website of business, e-commerce website as C2C (individual is to the individual) form, and the e-commerce website of B2C (businessman is to the individual) form, and can only buy a kind of e-commerce website of product for place an order in single businessman at every turn, " single purchase " concept is too narrow.Usually, this electron-like business web site operates as " single purchase " once placing an order, and the required product of user often need to could be bought complete in different businessmans by repeatedly placing an order, therefore, to once place an order and carry out the foundation of affairs set as " single purchase ", can be breaking at the relevance between the product that different businessmans buy, thereby affect the accuracy of final recommendation results.
On the other hand, a large amount of hot products are owing to having the purchase ubiquity, the a large amount of existence of the product mix (two product mixes, three product mixes etc..) that comprises these products have been caused, and it exists probability much larger than the product mix of other types, thereby makes the corresponding increase of the recommended probability of these product mixes.And the recommended ineffectivity that two aspects are arranged of hot product: the one, hot product is because public praise passes on from one to another or the reason such as ranking list, had highlyer by cognitive rate in customers, and the user need not by recommending just can to make to it decision-making of whether buying easily; The 2nd, what absolute support characterized is only the purchased number of times of product, what embody is the abundant sex character of statistics, can not embody the relevance between product, therefore can there be such situation: although it is very high to comprise the absolute support of product mix of hot product, but between the hot product that comprises in product mix and other products and uncorrelated or correlation degree is very low, say further, the highest support does not represent the highest degree of association, therefore, determine best collocation composition of relations based on the highest support, its accuracy remains to be discussed.
secondly, when selecting two product mixes or three product mixes to arrange in pairs or groups composition of relations as final the best, existing program is: as long as three product mix collection are not just determined PA according to three product mix set for empty entirely, in general, select the advantage of three product mix set to be suitably to expand the quantity that forms product, the user is had effective interest to expand, yet, when the correlation degree between product in three product mix set far below two product mix set in correlation degree between product, perhaps, when the correlation degree during three interior parts of product mix set make up between product is low, three product mix collection of simple selection are as net result, its accuracy remains to be discussed equally.
Summary of the invention
Method and device that the embodiment of the present invention provides the degree of association between a kind of product-based to determine the product mix relation in order to when the user chooses a certain product, improve to the degree of accuracy of the collocation product of its recommendation.
The concrete technical scheme that the embodiment of the present invention provides is as follows:
Between a kind of product-based, the degree of association provides the method for collocation product, comprising:
Receive user's selection request, determine target product corresponding to this selection request;
Determine the collocation product of described target product according to the collocation relation between product;
Return to described target product and the product of arranging in pairs or groups accordingly to the user;
Wherein, the collocation relation between described product obtains by following steps:
Obtain user's historical transaction record, and determine to set according to this historical transaction record the affairs set that forms in duration, wherein, any transaction table is taken over the set of all products of buying at the family for use in above-mentioned setting duration;
Form the product individual event collection of each product, a kind of product of each product individual event set representations, and calculate the absolute support of each product individual event collection, wherein, the absolute support of any one single product set is the number that comprises the affairs of this any one product;
Form the product N item collection of each product, N＞1, the combination of a N kind product of each product N item set representations, and calculate the absolute support of each product N item collection, wherein, the absolute support of any one product N item collection is the number that comprises the affairs of this any one product N item collection;
According to the absolute support of each product individual event collection that obtains and the absolute support of each product N item collection, calculate respectively the product degree of association of the product mix of each product N item set representations;
According to the product degree of association that each product N item set pair is answered, filter out respectively for each product and meet accordingly pre-conditioned product mix mode, and according to the final product mix mode of determining, determine the collocation product of each product.
Between a kind of product-based, the degree of association provides the device of collocation product, comprising:
Receiving element is used for receiving user's selection request, determines target product corresponding to this selection request;
Transmitting element is used for determining the collocation product of described target product according to the collocation relation between product, and returns to described target product and the product of arranging in pairs or groups accordingly to the user;
Wherein, the collocation relation between described product obtains by the following functions unit:
Acquiring unit is used for obtaining user's historical transaction record, and determines to set according to this historical transaction record the affairs set that forms in duration, and wherein, any transaction table is taken over the set of all products of buying at the family for use in above-mentioned setting duration;
The first processing unit, be used to form the product individual event collection of each product, a kind of product of each product individual event set representations, and calculate the absolute support of each product individual event collection, wherein, the absolute support of any one single product set is the number that comprises the affairs of this any one product; And the product N item collection that is used to form each product, N＞1, the combination of a N kind product of each product N item set representations, and calculate the absolute support of each product N item collection, wherein, the absolute support of any one product N item collection is the number that comprises the affairs of this any one product N item collection;
The second processing unit is used for calculating respectively the product degree of association of the product mix of each product N item set representations according to the absolute support of each product individual event collection that obtains and the absolute support of each product N item collection;
Determining unit is used for the product degree of association of answering according to each product N item set pair, filters out respectively for each product to meet accordingly pre-conditioned product mix mode, and according to the final product mix mode of determining, determines the collocation product of each product.
In the embodiment of the present invention, use distributed hardware, operating system and corresponding parallel program development language, realized following application function in electronic commerce network: at first, determine the affairs set based on setting duration, like this, guarantee on the one hand the validity that affairs form, also guaranteed on the other hand the relevance between product that affairs inside comprises; Secondly, when the absolute support of counting yield individual event collection and product N item collection, introduce correlativity and estimated the statistical property defective that makes up absolute support, thereby when the product-based degree of association determines that the best collocation between product concerns, can effectively improve the accuracy of determined collocation product, like this, when the user chooses a certain product, e-commerce website can recommend to arrange in pairs or groups more accurately product to it in time based on user's request, and then effectively guaranteed the E-business service quality, optimized the performance of e-commerce website.
Description of drawings
Fig. 1 is network architecture schematic diagram in the embodiment of the present invention;
Fig. 2 is that in the embodiment of the present invention, product mix concerns the computing platform illustrative view of functional configuration;
Fig. 3 is that in the embodiment of the present invention, product mix concerns that computing platform determines based on the degree of association between each product the collocation goods batch figure that each product is corresponding.
Embodiment
Defective in view of prior art, for the degree of association between can product-based, accurately judge the collocation relation between product, in the embodiment of the present invention, when the collocation between definite product concerns, employing setting duration (as, giving a discount one season, one between active stage etc.) interior user's consumer behavior determines the affairs set, like this, efficiently solve on the one hand the formation of affairs, also improved on the other hand the potential probability that has relevance between the product that affairs inside comprises; Secondly, on the basis of absolute support, introduce correlativity estimate (as, cosine measure), represent the relevance between product, better, in the present embodiment, any 2 parameters are defined as the degree of association (take cosine measure as example) of parameter A and parameter B:
(the affairs set quantity that contains parameter A and parameter B)/((the affairs set quantity that contains parameter A) * contains the affairs set quantity of parameter B))
1/2
Below in conjunction with accompanying drawing, the preferred embodiment of the present invention is elaborated.
Consult shown in Figure 1ly, in the embodiment of the present invention, the best collocation relation of user terminal, Web server, transaction data base server, product best collocation relation retrieve server and off-line product that includes in electronic commerce network is found computing platform.wherein, the user is mutual by user browser and e-commerce website that user terminal presents, transmission is browsed, search, buy request and relevant information, user's request receives by the Web server of e-commerce website the line correlation processing of going forward side by side, user's purchase information sends to the transaction data base service memory through Web server, the user of transaction data base server stores buys information and enters the best collocation of off-line product through the pre-service unloading and concern computing platform, the best collocation of this off-line product computing platform is by the distributed hardware cluster, Hadoop distributed operation environment and file system and form based on the Java language program of MapReduce.The best collocation of off-line product concerns that the computing platform regular update calculates the best collocation relation of product in the website, and produces the best matched combined of product on each line, calculates result of calculation to be updated to product the best relation retrieve server of arranging in pairs or groups after complete.When Web server receives the page request that comprises the best matched combined recommendation function of product, can send corresponding retrieval request to the best collocation of product relation retrieve server, the best collocation of product relation retrieve server is accepted this request, database on retrieval server, and with result feedback to Web server, by it, result is embedded in the final page and feeds back to user terminal, present to relative users by user terminal by user browser.
On the final line of realizing of the present invention, function is completed by said system, in actual applications, the best collocation of off-line product concerns that computing platform and the best collocation of product relation retrieve server can be separate functional entitys, it can be also functional modules different in same functional entity, concrete set-up mode is determined according to the complexity of actual application environment, is not repeated them here.
Consult shown in Figure 2ly, in the embodiment of the present invention, the best collocation of off-line product concerns that computing platform comprises receiving element 200, transmitting element 201, acquiring unit 20, the first processing unit 21, the second processing unit 22 and determining unit 23, wherein,
Receiving element 200 is used for receiving user's selection request, determines target product corresponding to this selection request;
Transmitting element 201 is used for determining the collocation product of target product according to the collocation relation between product, and returns to target product and the product of arranging in pairs or groups accordingly to the user;
Wherein, the collocation relation between the said goods obtains by the following functions unit:
Acquiring unit 20 is used for obtaining user's historical transaction record, and determines to set according to this historical transaction record the affairs set that forms in duration, and wherein, any transaction table is taken over the set of all products of buying at the family for use in above-mentioned setting duration;
The first processing unit 21, be used to form the product individual event collection of each product, a kind of product of each product individual event set representations, and calculate the absolute support of each product individual event collection, wherein, the absolute support of any one single product set is the number that comprises the affairs of this any one product; And the product N item collection that is used to form each product, N＞1, the combination of a N kind product of each product N item set representations, and calculate the absolute support of each product N item collection, wherein, the absolute support of any one product N item collection is the number that comprises the affairs of this any one product N item collection;
The second processing unit 22 is used for calculating respectively the product degree of association of the product mix of each product N item set representations according to the absolute support of each product individual event collection that obtains and the absolute support of each product N item collection;
Determining unit 23 is used for the product degree of association of answering according to each product N item set pair, filters out respectively for each product to meet accordingly pre-conditioned product mix mode, and according to the final product mix mode of determining, determines the collocation product of each product.
in actual applications, the best collocation of off-line product concerns that computing platform and the best collocation of product relation retrieve server might be separate devices, it might be also integrated device, if latter event, the best collocation of off-line product concerns that computing platform also can be on behalf of the function of completing the best collocation of product relation retrieve server, for example, consult shown in Figure 2, the best collocation of off-line product concern computing platform receive that Web server sends comprise the page request of the best matched combined recommendation function of product the time, database on retrieval server, the collocation product feedback that any products of Web server request is corresponding is to Web server, by it, result is embedded in the final page, and feed back to user terminal.
In following examples, for convenience of description, the best collocation of off-line product is concerned that computing platform concerns computing platform referred to as product mix, in the present embodiment, the best collocation relation of so-called product can be that the degree of association reaches the multiple product matched combined of setting threshold value, and a kind of collocation mode is not necessarily only arranged.
In practical application, product mix concerns that computing platform is when receiving user's selection request, first determine the target product that this selects the request correspondence, determine again the collocation product of this target product according to the collocation relation between product, and return to this target product and the product of arranging in pairs or groups accordingly to the user, wherein, the collocation relation between product concerns that by product mix between the computing platform product-based, the degree of association obtains, specifically consult shown in Figure 3ly, determine between product that the detailed process of collocation relation is as follows:
Step 300: product mix concerns the historical transaction record (as, each user's product purchased record in a year) of computing platform retrieval user in the transaction data base server.
In the present embodiment, product mix need to therefrom extract the purchase information that can be used for the counting yield degree of association after concerning that computing platform obtains user's historical transaction record, and finally forming corresponding tables of data, its pattern is＜user identifier, the time buying, product identifiers 〉, show referred to as RE.
Step 301: product mix concerns computing platform according to the affairs set that forms in the definite setting of the historical transaction record that obtains duration, and wherein, any transaction table is taken over the set of all products of buying at the family for use in above-mentioned setting duration.
For example: based on the RE table that generates in step 300, with a season as setting duration, the product set that a user was bought in a season forms affairs, one " Transaction Identifier " expresses by { user ID, season identifies }, and finally forms corresponding tables of data, its pattern is＜user identifier, season identifies, product identifiers 〉, show referred to as TP.
Step 302: product mix concerns that computing platform forms the product individual event collection of each product, a kind of product of each product individual event set representations, and calculate the absolute support of each product individual event collection, wherein, the absolute support of any one product individual event collection is the number that comprises the affairs of this any one product.
Be specially: each part product that the user was bought is respectively as a product individual event collection, and calculate respectively the absolute support of each product individual event collection, as, in the TP table, with the value of arbitrary " product identifiers " as a product individual event collection, and comprise the number of transactions (number of different { user ID, season identifies } value) of this product individual event collection as the absolute support of this product individual event collection.
The data pattern of result of calculation storage be＜product identifiers, absolute support 〉, show referred to as OneIAS.
Step 303: product mix concerns that computing platform forms the product N item collection combination of each product, N＞1, the combination of a N kind product of each product N item set representations, and calculate the absolute support of each N item product set, wherein, the absolute support of any one product N item collection is the number that comprises the affairs of this any one product N item collection.
In practical application, recommended requirements according to the user, when generating the product N item collection of each product, can generate corresponding product binomial collection, three collection of product, four collection of product, five collection of product ..., but in view of needs are controlled at algorithm complex to a certain degree interior to guarantee system performance, in the present embodiment, better, only in the mode that generates product binomial collection and three collection of product, the concrete enforcement of step 303 is introduced (but being not limited to this):
(1) at first, form the product binomial collection of each product, i.e. every product binomial collection of two product formations that is comprised by affairs at least.Transaction Identifier that the formation of product binomial collection is based on record in TP table calculates from connecting, and the data pattern of its result of calculation storage is＜product identifiers _ 1, product identifiers _ 2 〉, show referred to as TwoI.It is pointed out that the semantic repetition of data and set in the TwoI table, better, need pass through pre-conditioned screening from connection result, with the value of assurance product identifiers _ 1 less than the value of product identifiers _ 2 and the difference of each row of data.
Then, calculate the absolute support of each product binomial collection, the absolute support of any one product binomial collection is the number that comprises the affairs of this product binomial collection, can by product binomial collection corresponding to arbitrary row in retrieval TwoI table, calculate its absolute support.
The data pattern of its result of calculation storage be＜product identifiers _ 1, product identifiers _ 2, absolute support 〉, show referred to as TwoIAS.
(3) secondly, form three collection of product of each product, i.e. every three product mixes that comprised by affairs at least form three collection of a product.Transaction Identifier that the formation of three collection of product is based on record in TP table calculates from connecting, and the data pattern of its result of calculation storage is＜product identifiers _ 1, product identifiers _ 2, product identifiers _ 3 〉, show referred to as ThreeI.It is to be noted, avoid data and the repetition of gathering semanteme in the ThreeI table, better, need pass through pre-conditioned screening from connection result, with the value that guarantees product identifiers _ 1 value less than product identifiers _ 2, the value of product identifiers _ 2 is less than the value of product identifiers _ 3 and the difference of each row of data.
Then, calculate the absolute support of three collection of each product, the absolute support of three collection of any one product is the number that comprises the affairs of three collection of this product, can by three collection of product corresponding to arbitrary row in retrieval ThreeI table, calculate its absolute support.
The data pattern of the storage of its result of calculation is＜product identifiers _ 1, product identifiers _ 2, product identifiers _ 3, definitely support 〉, show referred to as ThreeIAS.
product individual event collection based on above-mentioned steps formation, three collection of product binomial collection and product, in practical application, in order to optimize result of calculation, better, in step 302, after obtaining the OneIAS table, also need further each product individual event collection of its record is screened, as, the absolute support that any product individual event set pair is answered, comprising the absolute support threshold value (setting in advance) of product binomial collection of this product individual event collection and the absolute support threshold value (setting in advance) that comprises three collection of product of this product individual event collection compares, if the value of the absolute support of above-mentioned any product individual event collection is minimum, with the deletion from the OneIAS table of this product individual event collection, on the other hand, also need the absolute support threshold value according to default product binomial collection, the value of absolute support is deleted from the TwoIAS table less than the product binomial collection of this threshold value, and, according to the absolute support threshold value of default three collection of product, the value of absolute support is deleted from the ThreeIAS table less than three collection of product of this threshold value.Like this, just can filter out and meet pre-conditioned product individual event collection, product binomial collection and three collection of product, thus convenient follow-up calculating operation.
Step 304: product mix concerns that computing platform according to the absolute support of each product individual event collection that obtains and the absolute support of each product N item collection, calculates respectively the product degree of association of the product mix of each product N item set representations.
Be specially:
(1) calculate the product degree of association of the product mix of each product binomial set representations.
For example, according to the information of record in record in OneIAS table and TwoIAS table, for arbitrary row＜product identifiers _ 1 in the TwoIAS table, product identifiers _ 2, absolute support 〉, calculate the degree of association of two kinds of products by following formula:
(absolute support/((the absolute support of Product Identifying 1) * (the absolute support of Product Identifying 2)) of product identifiers 1 and product identifiers 2 set
1/2
Wherein, the absolute support of product identifiers 1 and the absolute support of product identifiers 2 can obtain by inquiry OneIAS table, and the absolute support of product identifiers 1 and product identifiers 2 set can obtain by inquiry TwoIAS.
The data pattern of its result storage be＜product identifiers _ 1, product identifiers _ 2, the degree of association 〉, be called for short TwoIASR and show.
Better, above-mentioned result also needs to do further Screening Treatment, and screening conditions are two default product mix degree of association threshold values, will be by deletion from TwoIASR shows less than the product binomial collection of this threshold value.
(2) calculate the product degree of association of the product mix of three set representations of each product.
For example, information according to record in record in OneIAS table, TwoIAS table and ThreeIAS table, for any one product A, according to ThreeIAS table, retrieve three collection of all products that comprise A, for three collection＜product identifiers _ 1 of any one product wherein, product identifiers _ 2, product identifiers _ 3, definitely support 〉, calculate the degree of association of three collection of this product and product A by following formula:
(product identifiers _ 1, product identifiers _ 2, the absolute support of product identifiers _ 3 set
/ (the absolute support of the binomial product mix after product A is removed in product identifiers _ 3 set for (the absolute support of product A) * product identifiers _ 1, product identifiers _ 2))
1/2
Wherein, product identifiers _ 1, product identifiers _ 2, the absolute support of product identifiers _ 3 set can obtain by inquiry ThreeIAS table, the absolute support of product A can obtain by inquiry OneIAS table, product identifiers _ 1, product identifiers _ 2, the absolute support that the binomial product mix after A is removed in product identifiers _ 3 set can obtain by inquiry TwoIAS table.
The data pattern of its result storage be＜Product Identifying, 1,3 collocation Product Identifying 2 of 3 collocation Product Identifyings, the degree of association 〉, be called for short ThreeIASR and show.
Better, above-mentioned result also needs to do further Screening Treatment, and screening conditions are three default product mix degree of association threshold values, will be by deletion from ThreeIASR shows less than three collection of product of this threshold value.
Step 305: product mix concerns the product degree of association that computing platform is answered according to each product N item set pair, filters out respectively for each product to meet accordingly pre-conditioned product mix mode.
Be specially:
(1) according to the product binomial collection that forms, go out satisfactory product binomial collection (can for one or more) according to default binomial product screening conditional filtering, be about to it as the best collocation Candidate Set of binomial product mix.
For example, according to TwoIASR table, form binomial product mix table corresponding to each product, its data pattern be＜Product Identifying, 2 Product Identifyings of arranging in pairs or groups, the degree of association 〉, show referred to as TwoP.
Concrete calculating can realize by the mode with similar following SQL statement function:
Select product identifiers _ 1, product identifiers _ 2, the degree of association
From TwoIASR
Union all
Select product identifiers _ 2, product identifiers _ 1, the degree of association
From TwoIASR
Then, show according to TwoP, determine respectively in binomial product mix table corresponding to each product the most relevance degree (can be also time large degree of association or, most relevance degree and time large degree of association etc., herein only for for example), and the most relevance degree that each product is corresponding stores as result of calculation, its data pattern is＜Product Identifying, most relevance degree value 〉, show referred to as TwoMR; Concrete calculating can realize by the mode with similar following SQL statement function:
The Select Product Identifying, Max (degree of association)
From TwoP
Group by Product Identifying
At last, according to TwoP table and TwoMR table, calculate each product corresponding have a binomial product mix that maximum is answered the most relevance degree, and result of calculation is stored as the best collocation of binomial Candidate Set, its data pattern is＜Product Identifying, 2 best collocation candidate products signs 〉, show referred to as TwoPC.Because the value of the degree of association is possible identical, therefore, can comprise multiple binomial product mix in the best Candidate Set of arranging in pairs or groups of the binomial that any product is corresponding.
(2) according to three collection of product that form, go out three collection of satisfactory product (can for one or more) according to three default product screening conditional filterings, be about to its best as three product mixes Candidate Set of arranging in pairs or groups.
For example: show according to ThreeIASR, calculate most relevance degree in three product mixes corresponding to each product (can be also time large degree of association or, most relevance degree and time large degree of association etc., herein only for giving an example), and the most relevance degree that each product is corresponding is stored as result of calculation, its data pattern＜Product Identifying, most relevance degree value 〉, be called for short the ThreeMR table; Concrete calculating can realize by the mode with similar following SQL statement function:
The Select Product Identifying, Max (degree of association)
From ThreeIASR
Group by Product Identifying
Then, according to ThreeIASR table and ThreeMR table, calculate each product corresponding have three product mixes that maximum is answered the most relevance degree, and result of calculation is stored as three best collocation Candidate Sets, its data pattern is＜Product Identifying, 3 collocation Product Identifyings 1,3 collocation Product Identifyings 2 〉, be called for short the ThreePC table, because the value of the degree of association is possible identical, therefore, can comprise multiple three product mixes in three best collocation Candidate Sets that any product is corresponding.
Step 306: product mix concerns that computing platform according to the final product mix mode of determining, determines the collocation product of each product.
Be specially: for any product A, when determining its desirable collocation product, by retrieval TwoMR and ThreeMR, obtain respectively the best collocation of the binomial Candidate Set and three degrees of association that best collocation Candidate Set is corresponding of A; Then, carry out following judgement:
If the degree of association of the degree of association of three the best collocation Candidate Sets 〉=best collocation of binomial Candidate Set, select in three product mix set (hereinafter referred to as M3) corresponding to product A in ThreePC any combination as the best collocation relation of product A, and determine according to this collocation relation the collocation product that product A is corresponding.
If the degree of association of the degree of association of three the best collocation Candidate Sets＜best collocation of binomial Candidate Set is further processed binomial product mix set (hereinafter referred to as M2) corresponding to product A in the best collocation of binomial Candidate Set TwoPC, be specially:
If be the subset of M3 without combination in M2, select some binomial product mixes in M2 as the best relation of arranging in pairs or groups, with the final collocation product of determining product A, for example, can choose at random a binomial product mix from M2, perhaps, further choose a binomial product mix according to other conditions, as final selection.
If it is the subset of certain combination in M3 that combination is arranged in M2, select in M3 corresponding some three product mixes as the best relation of arranging in pairs or groups, with the final collocation product of determining product A.For example, suppose to have in M3 the subset of a plurality of combinations to appear in M2, choose at random in these combinations three product mixes, perhaps, further choose a binomial product mix according to other conditions, as final selection.
Certainly, in the present embodiment, in order to save calculated amount, under the prerequisite that meets user's request for utilization, can also by the mode of counting yield binomial collection only or three collection of product, judge the best collocation relation of each product, in the present embodiment, only describe as example with while counting yield binomial collection and three modes that integrate of product, do not repeat them here.
Based on above-described embodiment, in practical application, product mix concerns that computing platform can be after the selection request that receives user's transmission and determining corresponding target product, execution in step 300-step 306 obtains the collocation relation between product, then, determine again the collocation product of target product according to the collocation relation between acquired product, and return to target product and the product of arranging in pairs or groups accordingly to the user; Perhaps, product mix concern computing platform also in advance execution in step 300-step 306 obtain collocation relation between product, and determine corresponding target product after receiving the selection request that the user sends, and the collocation product of determining target product according to the collocation relation between acquired product, and return to target product and the product of arranging in pairs or groups accordingly to the user.I.e. expression, between product, can arrange flexibly the opportunity of determining of collocation relation according to the practical application needs, is not limited to a kind of implementation pattern, except above-mentioned two kinds of executive modes, also has other multiple set-up modes flexibly, do not repeat them here.
On the other hand, in the above-described embodiments, the degree of association between product is based on the cosine test and calculates, and in actual applications, can also adopt the statistical measurement compute associations degree of other types:
For example, can adopt coverage (coverage) to come the degree of association between counting yield as statistical test, coverage is defined as follows:
Wherein, set Y is expressed as with the coverage of product X: the absolute support of the set that X becomes with Y shape is divided by the absolute support of Y.
Again for example, can also adopt lifting degree (lift) to come the degree of association between counting yield as statistical measurement, the lifting degree is defined as follows:
Wherein, set Y is shown with the lifting kilsyth basalt of product X: the absolute support of the set that X becomes with Y shape is divided by the product of the absolute support of the absolute support of Y and X.
In sum, in the embodiment of the present invention, use distributed hardware, operating system and corresponding parallel program development language, realized following application function in electronic commerce network: at first, determine the affairs set based on setting duration, like this, guarantee on the one hand the validity that affairs form, also guaranteed on the other hand the relevance between product that affairs inside comprises; Secondly, when the absolute support of counting yield individual event collection and product N item collection, introduce correlativity and estimated the statistical property defective that makes up absolute support, thereby when the product-based degree of association determines that the best collocation between product concerns, can effectively improve the accuracy of determined collocation product, like this, when the user chooses a certain product, e-commerce website can recommend to arrange in pairs or groups more accurately product to it in time based on user's request, and then effectively guaranteed the E-business service quality, optimized the performance of e-commerce website.
Further, in the present embodiment, finally to determine the best collocation relation between product by considering set-inclusion relation between the multiple product array mode, like this, just from overall candidate space and local candidate two, space angle, the definite of final best collocation relation introduced the stricter mechanism of selecting, thereby further promoted the accuracy of the final product mix relation of determining.
Obviously, those skilled in the art can carry out various changes and modification and not break away from the spirit and scope of the present invention the present invention.Like this, if within of the present invention these are revised and modification belongs to the scope of claim of the present invention and equivalent technologies thereof, the present invention also is intended to comprise these changes and modification interior.
Claims (10)
1. between a product-based, the degree of association provides the method for collocation product, it is characterized in that, comprising:
Receive user's selection request, determine target product corresponding to this selection request;
Determine the collocation product of described target product according to the collocation relation between product;
Return to described target product and the product of arranging in pairs or groups accordingly to the user;
Wherein, the collocation relation between described product obtains by following steps:
Obtain user's historical transaction record, and determine to set according to this historical transaction record the affairs set that forms in duration, wherein, any transaction table is taken over the set of all products of buying at the family for use in above-mentioned setting duration;
Form the product individual event collection of each product, a kind of product of each product individual event set representations, and calculate the absolute support of each product individual event collection, wherein, the absolute support of any one single product set is the number that comprises the affairs of this any one product;
Form the product N item collection of each product, N＞1, the combination of a N kind product of each product N item set representations, and calculate the absolute support of each product N item collection, wherein, the absolute support of any one product N item collection is the number that comprises the affairs of this any one product N item collection;
According to the absolute support of each product individual event collection that obtains and the absolute support of each product N item collection, calculate respectively the product degree of association of the product mix of each product N item set representations;
According to the product degree of association that each product N item set pair is answered, filter out respectively for each product and meet accordingly pre-conditioned product mix mode, and according to the final product mix mode of determining, determine the collocation product of each product.
2. the method for claim 1, it is characterized in that, form the product N item collection of each product, N＞1, the combination of a N kind product of each product N item set representations, and calculate the absolute support of each product N item set, wherein, the absolute support of any one product N item collection is the number that comprises the affairs of this any one product N item collection, comprising:
Form the product binomial collection of each product, the combination of two kinds of products of each product binomial set representations, and calculate the absolute support of each product binomial collection, wherein, the absolute support of any one product binomial collection is the number that comprises the affairs of this any one product binomial collection;
Form three collection of product of each product, the combination of three kinds of products of three set representations of each product, and calculate the absolute support of three collection of each product, wherein, the absolute support of three collection of any one product is the number that comprises the affairs of three collection of this any one product.
3. the method for claim 1, is characterized in that, in the absolute support that obtains each product individual event set pair and answer, and after each product N item set pair absolute support of answering, further comprises:
Absolute support is not met pre-conditioned product individual event collection deletes; And,
The product N item collection that absolute support is not reached default absolute support threshold value is deleted.
4. method as claimed in claim 2, is characterized in that, according to the absolute support of each product individual event collection that obtains and the absolute support of each product N item collection, calculates respectively the product degree of association of the product mix of any one product N item set representations, comprising:
If described any one product N item integrates as product binomial collection, the absolute support of answering according to this product binomial set pair, and each product individual event collection that this product binomial collection comprises is distinguished corresponding absolute support, the degree of association of the product mix of the described product binomial set representations of calculating acquisition;
If described any one product N item integrates as three collection of product, the absolute support of answering according to three set pairs of this product, the absolute support of the product individual event collection that appointed product characterizes, and three of described products concentrate and remove definitely support of two set pairs of product of characterizing after described appointed product, calculate the degree of association between acquisition three collection of described product and described appointed product.
5. method as claimed in claim 4, is characterized in that, after the product degree of association of the product mix of determining each product N item set representations, further comprises:
The product N item collection that the product degree of association is not reached default degree of association threshold value is deleted.
6. according to claim 2-5 described methods of any one, is characterized in that, according to the product degree of association that each product N item set pair is answered, filter out any product corresponding meet pre-conditioned product mix mode, comprising:
According to the product degree of association that each product binomial set pair that comprises described any product is answered, filter out the product degree of association and reach the product binomial collection of setting threshold value, and determine the represented binomial product mix mode set of product binomial collection chosen;
According to the product degree of association between three collection of each product that comprise described any product and described any product, filter out the product degree of association and reach three collection of product of setting threshold value, and determine that three of the products of choosing collect represented three product mix modes set.
7. method as claimed in claim 6, is characterized in that, according to the final product mix mode of determining, determines the collocation product of any product, comprising:
Judge three product mix mode set expressions that described any product is corresponding and described any product between the product degree of association, whether more than or equal to the product degree of association of binomial product mix mode set expression corresponding to described any product;
If select a kind of three product mix modes to determine the collocation product of described any product in the set of described three product mix modes according to predetermined manner;
Otherwise, further judgement is in binomial product mix mode set and the set of three product mix modes, whether having a certain binomial product mix mode is the subset of a certain three product mix modes set, if determine the collocation product of described any product according to described a certain three product mix modes; Otherwise, select a kind of binomial product mix mode in the set of described binomial product mix mode according to predetermined manner, determine the collocation product of described any product.
8. as the described method of claim 2-5 any one, it is characterized in that the step of the collocation relation between described acquisition product, carried out before the selection request that receives the user, perhaps, after the selection request that receives the user, carry out in the process of the collocation product of determining described target product.
9. between a product-based, the degree of association provides the device of collocation product, it is characterized in that, comprising:
Receiving element is used for receiving user's selection request, determines target product corresponding to this selection request;
Transmitting element is used for determining the collocation product of described target product according to the collocation relation between product, and returns to described target product and the product of arranging in pairs or groups accordingly to the user;
Wherein, the collocation relation between described product obtains by the following functions unit:
Acquiring unit is used for obtaining user's historical transaction record, and determines to set according to this historical transaction record the affairs set that forms in duration, and wherein, any transaction table is taken over the set of all products of buying at the family for use in above-mentioned setting duration;
The first processing unit, be used to form the product individual event collection of each product, a kind of product of each product individual event set representations, and calculate the absolute support of each product individual event collection, wherein, the absolute support of any one single product set is the number that comprises the affairs of this any one product; And the product N item collection that is used to form each product, N＞1, the combination of a N kind product of each product N item set representations, and calculate the absolute support of each product N item collection, wherein, the absolute support of any one product N item collection is the number that comprises the affairs of this any one product N item collection;
The second processing unit is used for calculating respectively the product degree of association of the product mix of each product N item set representations according to the absolute support of each product individual event collection that obtains and the absolute support of each product N item collection;
Determining unit is used for the product degree of association of answering according to each product N item set pair, filters out respectively for each product to meet accordingly pre-conditioned product mix mode, and according to the final product mix mode of determining, determines the collocation product of each product.
10. device as claimed in claim 9, it is characterized in that, first by the collocation relation between described acquiring unit, the first processing unit, the second processing unit and determining unit acquisition product, receive again user's selection request by described receiving element, perhaps, after described receiving element receives user's selection request, determine in the process of collocation product of described target product at described transmitting element, obtain collocation relation between product by described acquiring unit, the first processing unit, the second processing unit and determining unit.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201110343441.8A CN103093369B (en) | 2011-11-03 | A kind of method and device that collocation product is provided based on the degree of association between product |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201110343441.8A CN103093369B (en) | 2011-11-03 | A kind of method and device that collocation product is provided based on the degree of association between product |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN103093369A true CN103093369A (en) | 2013-05-08 |
| CN103093369B CN103093369B (en) | 2016-12-14 |
Family
ID=
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105205684A (en) * | 2014-06-30 | 2015-12-30 | 阿里巴巴集团控股有限公司 | Recommended display method of matched products and apparatus |
| CN106096991A (en) * | 2016-05-19 | 2016-11-09 | 北京三快在线科技有限公司 | The automatic generation method of a kind of set meal and device |
| WO2016188277A1 (en) * | 2015-05-25 | 2016-12-01 | 阿里巴巴集团控股有限公司 | Method and device for providing business object collocation information |
| WO2016188278A1 (en) * | 2015-05-25 | 2016-12-01 | 阿里巴巴集团控股有限公司 | Method and device for collecting information about service object |
| CN107146127A (en) * | 2017-03-22 | 2017-09-08 | 重庆允升科技有限公司 | A kind of industrial goods accurate recommendation method and system |
| CN107256513A (en) * | 2017-06-19 | 2017-10-17 | 数据堂（北京）科技股份有限公司 | Method and device is recommended in a kind of collocation of object |
| CN107870936A (en) * | 2016-09-27 | 2018-04-03 | 腾讯科技（深圳）有限公司 | The related effective item set mining method, apparatus of data item and data processing equipment |
| CN108320213A (en) * | 2018-01-31 | 2018-07-24 | 深圳春沐源控股有限公司 | Electric business Method of Commodity Recommendation and electric business Platform Server |
| CN111563782A (en) * | 2019-02-13 | 2020-08-21 | 北京京东尚科信息技术有限公司 | Method and terminal for determining product to be recommended |
| CN111639974A (en) * | 2020-06-02 | 2020-09-08 | 海汇星驰信息科技(广州)有限公司 | Product association degree quantitative analysis method based on amazon platform |
| CN112634009A (en) * | 2020-12-14 | 2021-04-09 | 中国建设银行股份有限公司 | Product pushing method and device, storage medium and electronic equipment |
| US11538084B2 (en) | 2018-12-14 | 2022-12-27 | Hewlett Packard Enterprise Development Lp | Method and medium for customer product recommendations |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101206751A (en) * | 2007-12-25 | 2008-06-25 | 北京科文书业信息技术有限公司 | Customer recommendation system based on data digging and method thereof |
| CN101324948A (en) * | 2008-07-24 | 2008-12-17 | 阿里巴巴集团控股有限公司 | Method and apparatus of recommending information |
| US20100088148A1 (en) * | 2008-10-02 | 2010-04-08 | Presswala Irfan | System and methodology for recommending purchases for a shopping intent |
| CN101799810A (en) * | 2009-02-06 | 2010-08-11 | 中国移动通信集团公司 | Association rule mining method and system thereof |
| CN102073717A (en) * | 2011-01-07 | 2011-05-25 | 南京大学 | Home page recommending method for orienting vertical e-commerce website |
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101206751A (en) * | 2007-12-25 | 2008-06-25 | 北京科文书业信息技术有限公司 | Customer recommendation system based on data digging and method thereof |
| CN101324948A (en) * | 2008-07-24 | 2008-12-17 | 阿里巴巴集团控股有限公司 | Method and apparatus of recommending information |
| US20100088148A1 (en) * | 2008-10-02 | 2010-04-08 | Presswala Irfan | System and methodology for recommending purchases for a shopping intent |
| CN101799810A (en) * | 2009-02-06 | 2010-08-11 | 中国移动通信集团公司 | Association rule mining method and system thereof |
| CN102073717A (en) * | 2011-01-07 | 2011-05-25 | 南京大学 | Home page recommending method for orienting vertical e-commerce website |
Non-Patent Citations (1)
| Title |
|---|
| 薛洁，刘希玉: "数据挖掘技术与网上购物推荐系统", 《网络安全技术与应用》, no. 03, 15 March 2011 (2011-03-15), pages 58 - 60 * |
Cited By (19)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105205684A (en) * | 2014-06-30 | 2015-12-30 | 阿里巴巴集团控股有限公司 | Recommended display method of matched products and apparatus |
| WO2016188277A1 (en) * | 2015-05-25 | 2016-12-01 | 阿里巴巴集团控股有限公司 | Method and device for providing business object collocation information |
| WO2016188278A1 (en) * | 2015-05-25 | 2016-12-01 | 阿里巴巴集团控股有限公司 | Method and device for collecting information about service object |
| CN106294420A (en) * | 2015-05-25 | 2017-01-04 | 阿里巴巴集团控股有限公司 | The method and device of business object collocation information is provided |
| CN106294419A (en) * | 2015-05-25 | 2017-01-04 | 阿里巴巴集团控股有限公司 | The collecting method of business object information and device |
| CN106294420B (en) * | 2015-05-25 | 2019-11-05 | 阿里巴巴集团控股有限公司 | The method and device of business object collocation information is provided |
| JP2018515861A (en) * | 2015-05-25 | 2018-06-14 | アリババ・グループ・ホールディング・リミテッドＡｌｉｂａｂａ Ｇｒｏｕｐ Ｈｏｌｄｉｎｇ Ｌｉｍｉｔｅｄ | Method and apparatus for providing collocation information for business objects |
| CN106096991A (en) * | 2016-05-19 | 2016-11-09 | 北京三快在线科技有限公司 | The automatic generation method of a kind of set meal and device |
| CN107870936A (en) * | 2016-09-27 | 2018-04-03 | 腾讯科技（深圳）有限公司 | The related effective item set mining method, apparatus of data item and data processing equipment |
| CN107146127A (en) * | 2017-03-22 | 2017-09-08 | 重庆允升科技有限公司 | A kind of industrial goods accurate recommendation method and system |
| CN107256513A (en) * | 2017-06-19 | 2017-10-17 | 数据堂（北京）科技股份有限公司 | Method and device is recommended in a kind of collocation of object |
| CN108320213A (en) * | 2018-01-31 | 2018-07-24 | 深圳春沐源控股有限公司 | Electric business Method of Commodity Recommendation and electric business Platform Server |
| US11538084B2 (en) | 2018-12-14 | 2022-12-27 | Hewlett Packard Enterprise Development Lp | Method and medium for customer product recommendations |
| CN111563782A (en) * | 2019-02-13 | 2020-08-21 | 北京京东尚科信息技术有限公司 | Method and terminal for determining product to be recommended |
| CN111563782B (en) * | 2019-02-13 | 2024-05-21 | 北京京东尚科信息技术有限公司 | Method and terminal for determining to-be-recommended products |
| CN111639974A (en) * | 2020-06-02 | 2020-09-08 | 海汇星驰信息科技(广州)有限公司 | Product association degree quantitative analysis method based on amazon platform |
| CN111639974B (en) * | 2020-06-02 | 2021-04-06 | 海汇星驰信息科技(广州)有限公司 | Product association degree quantitative analysis method based on amazon platform |
| CN112634009A (en) * | 2020-12-14 | 2021-04-09 | 中国建设银行股份有限公司 | Product pushing method and device, storage medium and electronic equipment |
| CN112634009B (en) * | 2020-12-14 | 2024-05-14 | 中国建设银行股份有限公司 | Product pushing method and device, storage medium and electronic equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US20190228323A1 (en) | Systems and methods for trend aware self-correcting entity relationship extraction | |
| CN102792299B (en) | Feeding specific to user is recommended | |
| US20190012683A1 (en) | Method for predicting purchase probability based on behavior sequence of user and apparatus for the same | |
| CN108664513B (en) | Method, device and equipment for pushing keywords | |
| KR102042047B1 (en) | System and method of marketing service using formal and informal big data | |
| US10380123B2 (en) | Store-specific recommendation engine | |
| CN103207881B (en) | Querying method and device | |
| JP6723182B2 (en) | Information processing apparatus, information processing method, and program | |
| US8626604B1 (en) | Aggregating product endorsement information | |
| WO2013025874A2 (en) | Page reporting | |
| JP2012523643A (en) | Proposal of network object information for users | |
| CN103118111A (en) | Information push method based on data from a plurality of data interaction centers | |
| CN103995866A (en) | Commodity information pushing method and device based on link forecasting | |
| CN103678365A (en) | Method, device and system for dynamically obtaining data | |
| US9065827B1 (en) | Browser-based provisioning of quality metadata | |
| CN103885971A (en) | Data pushing method and data pushing device | |
| AU2016203941A1 (en) | Visitor identification based on feature selection | |
| JP7231317B2 (en) | Estimation device, estimation method and estimation program | |
| CN109190027A (en) | Multi-source recommended method, terminal, server, computer equipment, readable medium | |
| KR20130033506A (en) | Item recommendation method and apparatus | |
| JP2006053616A (en) | Server device, web site recommendation method and program | |
| CN109426998B (en) | Information pushing method and device | |
| CN110874639A (en) | Method and device for acquiring operation information | |
| CN103093369A (en) | Method and device for offering matched product based on correlation degree between products | |
| CN106557475B (en) | Method and device for selecting new path |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| REG | Reference to a national code |
Ref country code: HK Ref legal event code: DE Ref document number: 1180097 Country of ref document: HK |
|
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| REG | Reference to a national code |
Ref country code: HK Ref legal event code: GR Ref document number: 1180097 Country of ref document: HK |