CN110992153A - Commodity recommendation method, system and equipment based on user attributes and commodity types - Google Patents
Commodity recommendation method, system and equipment based on user attributes and commodity types Download PDFInfo
- Publication number
- CN110992153A CN110992153A CN201911304899.5A CN201911304899A CN110992153A CN 110992153 A CN110992153 A CN 110992153A CN 201911304899 A CN201911304899 A CN 201911304899A CN 110992153 A CN110992153 A CN 110992153A
- Authority
- CN
- China
- Prior art keywords
- commodity
- user
- frequency
- weighted
- weight
- 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
- 238000000034 method Methods 0.000 title claims abstract description 59
- 239000011159 matrix material Substances 0.000 claims description 70
- 238000010606 normalization Methods 0.000 claims description 21
- 238000004364 calculation method Methods 0.000 claims description 18
- 230000001960 triggered effect Effects 0.000 claims description 15
- 230000008569 process Effects 0.000 claims description 11
- 238000004590 computer program Methods 0.000 claims description 7
- 238000010276 construction Methods 0.000 claims description 2
- 238000010586 diagram Methods 0.000 description 6
- 238000011156 evaluation Methods 0.000 description 6
- 230000000694 effects Effects 0.000 description 5
- 230000006399 behavior Effects 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 2
- 238000001914 filtration Methods 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 238000007418 data mining Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000007670 refining Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/06—Buying, selling or leasing transactions
- G06Q30/0601—Electronic shopping [e-shopping]
- G06Q30/0631—Item recommendations
Abstract
The invention provides a commodity recommendation method based on user attributes and commodity types, which comprises the steps of firstly forming a user attribute-commodity type binary relation network by using attribute information of a user and type information of commodities purchased by the user, and then calculating the association weight, frequency weight, value weight, triggering centrality and network core degree of the commodities according to certain type of attribute information of the user; and then carrying out weighted summation on the weighted association weight, the weighted frequency weight, the weighted value weight, the weighted triggering centrality and the weighted network core degree to obtain the optimal node degree of the classified user commodities, and recommending the commodities to corresponding classified users according to the optimal node degree of the classified user commodities. The method can obtain the commodity which has the widest spread and the largest influence under a certain user attribute, and the commodity pushing is more detailed and accurate.
Description
Technical Field
The invention relates to the field of data mining, in particular to a commodity recommendation method, a commodity recommendation system and commodity recommendation equipment based on user attributes and commodity types.
Background
The commodity recommendation method in the prior art usually adopts a collaborative filtering recommendation technology or a point degree centrality technology of a relationship network.
The drawbacks of collaborative filtering recommendations are: (1) it depends on the recommendation; that is, the user can only buy a certain amount of goods (in practical application, the buying can also represent the favorite selection conditions such as clicking, liking, selecting, collecting, adding shopping cart, etc.), and then the user can have the recommended target, which belongs to passive recommendation. (2) The recommended goods only consider single purchase (namely the accurate association behavior based on only one time), but not consider secondary derivative purchase, and the subsequent derivative recommendation effect is poor.
The defects of the dot-degree centrality of the relationship network are as follows: (1) only the transmissibility of the commodity (whether the commodity is linked with many other commodities) is considered, but the influence of the commodity is not considered. Firstly, although the commodity is associated with a plurality of commodities, the proportion of the number of times of associated and simultaneous purchase of the commodity to the total number of times of purchase of the commodity is low, for example, 10 times of total purchase of the commodity A are performed, 9 times of purchase are single purchase, and only 1 time of simultaneous purchase of other commodities is performed; secondly, although the commodity is connected with a plurality of other commodities, the purchasing frequency of the commodity is low, the commodity is not popular, the influence is poor, the possibility that the commodity is not purchased by audiences at all is high, and the overall effect is poor. Thirdly, although the commodity is related to a plurality of commodities, the total value of the commodity is low, and no money is made for a platform operator; (2) the refining degree is insufficient: the general recommendation does not distinguish the gender, income and other user attribute characteristics of the user, and the effect of the general recommendation possibly affects the users with different attributes to different degrees, so that the recommendation effect generates larger difference.
Disclosure of Invention
The embodiment of the invention aims to provide a commodity recommendation method, a system and equipment based on user attributes and commodity types, and aims to solve the problems that in the prior art, commodity factors are not considered comprehensively, and attribute characteristics of a user are not considered, so that the recommendation effect is poor and inaccurate.
A first object of an embodiment of the present invention is to provide a method for recommending a product based on a user attribute and a product type, where the method includes:
establishing a classified user commodity association information table according to the user commodity transaction data;
counting the total purchase frequency of each user category to each commodity; counting the associated purchase frequency of each user category to each commodity;
calculating the total value of each commodity according to the total purchase frequency of each commodity and the unit price of each commodity of each user category;
calculating the association weight, frequency weight and value weight of each commodity of each user category;
counting the frequency of the commodities of each user category which are purchased simultaneously according to the commodity transaction data of the users; creating a directional commodity purchasing frequency matrix of each user category;
creating an undirected commodity purchasing frequency matrix of each user category according to the directed commodity purchasing frequency matrix of each user category;
carrying out normalization processing on the directional commodity purchasing frequency matrixes of all user categories to obtain corresponding directional commodity purchasing weighting frequency matrixes;
calculating the triggering centrality of each commodity of each user category;
calculating the network core degree of each commodity of each user category;
respectively carrying out normalization processing on the association weight, the frequency weight, the value weight, the triggering centrality and the network core degree of each commodity of each user category to obtain the corresponding weighted association weight, weighted frequency weight, weighted value weight, weighted triggering centrality and weighted network core degree;
weighting and summing the weighted association weight, the weighted frequency weight, the weighted value weight, the weighted triggering centrality and the weighted network core degree to obtain the optimal node degree for classifying the user commodities;
and recommending the commodities to the corresponding classified users according to the optimal node degree of the commodities of the classified users.
A second object of an embodiment of the present invention is to provide a product recommendation system based on user attributes and product types, the system including:
the user commodity associated information table creating module is used for creating a classified user commodity associated information table according to the user commodity transaction data;
the commodity purchase frequency counting module is used for counting the total purchase frequency of each commodity for each user category; counting the associated purchase frequency of each user category to each commodity;
the commodity total value calculating module is used for calculating the total value of each commodity according to the total purchase frequency of each user type to each commodity and the unit price of each commodity;
the weight calculation module is used for calculating the association weight, the frequency weight and the value weight of each commodity of each user category;
the oriented commodity purchase frequency matrix creating module is used for counting the oriented and simultaneous purchase frequency of commodities of various user categories according to the commodity transaction data of the users; creating a directional commodity purchasing frequency matrix of each user category;
the undirected commodity purchasing frequency matrix creating module is used for creating an undirected commodity purchasing frequency matrix of each user category according to the directed commodity purchasing frequency matrix of each user category;
the first normalization processing module is used for performing normalization processing on the directional commodity purchasing frequency matrixes of all user categories to obtain corresponding directional commodity purchasing weighting frequency matrixes;
the triggering centrality calculating module is used for calculating the triggering centrality of each commodity of each user category;
a network core degree calculation device for calculating the network core degree of each commodity of each user category;
the second normalization processing module is used for respectively normalizing the association weight, the frequency weight, the value weight, the triggering centrality and the network core degree of each commodity of each user category to obtain the corresponding weighted association weight, the weighted frequency weight, the weighted value weight, the weighted triggering centrality and the weighted network core degree;
the optimal node degree calculation module is used for carrying out weighted summation on the weighted association weight, the weighted frequency weight, the weighted value weight, the weighted triggering centrality and the weighted network core degree to obtain the optimal node degree of the classified user commodities;
and the commodity recommending module is used for recommending commodities to the corresponding classified users according to the optimal node degree of the commodities of the classified users.
It is a third object of an embodiment of the present invention to provide an apparatus, including a memory, a processor, and a computer program stored in the memory and executable on the processor, where the processor implements the steps of the product recommendation method based on the user attribute and the product type when executing the computer program.
The invention has the advantages of
The invention provides a commodity recommendation method based on user attributes and commodity types, which comprises the steps of firstly forming a user attribute-commodity type binary relation network by using attribute information of a user and type information of commodities purchased by the user, and then calculating the association weight, frequency weight, value weight, triggering centrality and network core degree of the commodities according to certain type of attribute information of the user; and then carrying out weighted summation on the weighted association weight, the weighted frequency weight, the weighted value weight, the weighted triggering centrality and the weighted network core degree to obtain the optimal node degree of the classified user commodities, and recommending the commodities to corresponding classified users according to the optimal node degree of the classified user commodities. The method can obtain the commodity which has the widest spread and the largest influence under a certain user attribute, and the commodity pushing is more detailed and accurate.
Drawings
Fig. 1 is a flowchart of a commodity recommendation method based on user attributes and commodity types according to an embodiment of the present invention;
fig. 2 is a flowchart of a method for calculating a network core degree of each commodity of each user category according to an embodiment of the present invention;
FIG. 3 is a commodity node G provided by the embodiment of the present invention1、G2、G3、G4、G5A connection relationship network diagram of (1);
fig. 4 is a structural diagram of a product recommendation system based on user attributes and product types according to an embodiment of the present invention;
fig. 5 is a structural diagram of a network core degree computing device according to an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is further described in detail below with reference to the accompanying drawings and examples, and for convenience of description, only parts related to the examples of the present invention are shown. It is to be understood that the specific embodiments described herein are for purposes of illustration only and not for purposes of limitation, as other equivalent embodiments may be devised in accordance with the embodiments of the present invention by those of ordinary skill in the art without the use of inventive faculty.
The invention provides a commodity recommendation method based on user attributes and commodity types, which comprises the steps of firstly forming a user attribute-commodity type binary relation network by using attribute information of a user and type information of commodities purchased by the user, and then calculating the association weight, frequency weight, value weight, triggering centrality and network core degree of the commodities according to certain type of attribute information of the user; and then carrying out weighted summation on the weighted association weight, the weighted frequency weight, the weighted value weight, the weighted triggering centrality and the weighted network core degree to obtain the optimal node degree of the classified user commodities, and recommending the commodities to corresponding classified users according to the optimal node degree of the classified user commodities. The method can obtain the commodity which has the widest spread and the largest influence under a certain user attribute, and the commodity pushing is more detailed and accurate.
Fig. 1 is a flowchart of a commodity recommendation method based on user attributes and commodity types according to an embodiment of the present invention; the method comprises the following steps:
step1, creating a classification user commodity association information table according to the user commodity transaction data;
the user commodity transaction data includes fields: user number, user category, commodity category and transaction time;
the classification user commodity association information table comprises the following fields: a user number, a user category and a plurality of specific commodity categories; when the value in the user commodity association information table indicates that one-time purchasing behavior occurs, whether the corresponding row user purchases the corresponding commodity or not is judged, 1 indicates purchasing, and 0 indicates not purchasing; each row in the table represents a purchase record; the same user can have multiple purchase records;
in a specific embodiment, the user number may be a number, a character, a combination thereof, or the like, which can uniquely mark the user identity, such as a mobile phone number, a card number, a serial number, or the like; the user category is classified according to user attributes or characteristics, such as gender, age, family income, region, and the like; the commodity transaction data is acquired from a business super or e-commerce platform and the like;
table 1 shows the user commercial transaction data segments according to the embodiment of the present invention; the user number is a card number, and the user category is gender; the commodity is a paid video, including a love paid video, a comprehensive art paid video and the like.
Card number | Sex | Commodity | Transaction time |
26630 | Woman | Love | 11-01 19:00 |
26630 | Woman | Comprehensive art | 11-01 19:20 |
26630 | Woman | Youth of youth | 11-01 19:40 |
62995 | Woman | Comprehensive art | 11-02 17:50 |
62995 | Woman | Cartoon | 11-02 19:00 |
38765 | For male | Movement of | 11-03 18:00 |
… | ... | ... | ... |
TABLE 1
In particular embodiments, the purchased item may also indicate a favorite selection, viewed, selected, collected, shopping cart added, and the like; the lack of purchase may also mean dislike, lack of viewing, or lack of selection, etc., which one skilled in the art would understand is not intended to limit the scope of the present invention;
as shown in table 2, the present invention classifies the user commodity associated information table segments;
TABLE 2
Step2, counting the total purchase frequency of each commodity of each user category; counting the associated purchase frequency of each user category to each commodity;
in the embodiment of the invention, the commodity platform is set to have n commodities in total, and the set G ═ G is used1,G2,….GnRepresents; n represents the number of commodity types, also called the number of commodity nodes; there are a total of k user categories, with the set C ═ C1,c2,….ckDenotes, i ═ 1,2, 3.. n }, and x ═ 1,2, 3.. k };
user class cxFor commodity GiIs given by k × n matrix CF3 { (R3 (c)x))iDenotes (R3 (c)x))iThe value of (D) represents the user class cxFor commodity GiTotal frequency of purchases;
user class cxFor commodity GiIs given by k × n matrix CF2 { (R2 (c)x))iDenotes (R2 (c)x))iThe value of (D) represents the user class cxFor commodity GiAssociated purchase frequency of;
(R2(cx))i=(R3(cx))i-(R1(cx))i(ii) a Wherein (R1 (c)x))iRepresenting a user class cxFor commodity GiIndividual purchase frequency of; commodity GiThe individual purchase frequency of (A) indicates that only a single commodity G is purchasediThe frequency of (2); commodity GiThe associated purchase frequency of (2) indicates that the purchased product includes product GiFrequency of (i.e. article G)iFrequency of purchases at the same time with all other merchandise;
as shown in table 3, a data table of total purchase frequency of commodities for male and female users according to the embodiment of the present invention;
love | Suspense questions of the present invention | Art and literature | Comprehensive art | Swordsman | Movement of | History of | Youth of youth | Crime | War | Cartoon | |
For male | 0 | 2 | 0 | 0 | 4 | 6 | 2 | 3 | 6 | 4 | 6 |
Woman | 6 | 1 | 4 | 18 | 2 | 2 | 1 | 10 | 2 | 0 | 2 |
TABLE 3
As shown in table 4, a data table of commodity association purchase frequency for male and female users according to the embodiment of the present invention;
love | Suspense questions of the present invention | Art and literature | Comprehensive art | Swordsman | Movement of | History of | Youth of youth | Crime | War | Cartoon | |
For male | 0 | 1 | 0 | 0 | 2 | 2 | 1 | 2 | 2 | 2 | 3 |
|
5 | 0 | 4 | 10 | 1 | 1 | 0 | 6 | 1 | 0 | 1 |
TABLE 4
Step3, calculating the total value of each commodity according to the total purchase frequency and the unit price of each commodity of each user type;
user class cxPurchased article GiThe total value of (a) is determined by k × n matrix CF4 { (Val3 (c)x))iDenotes (Val3 (c)x))iThe value of (D) represents the user class cxPurchased article GiThe total value of (c);
(Val3(cx))i=(R3(cx))i×(Pr)i;
as shown in table 5, a data table of total value of each commodity for the category of the male user according to the embodiment of the present invention;
love | Suspense questions of the present invention | Art and literature | Comprehensive art | Swordsman | Movement of | History of | Youth of youth | Crime | War | Cartoon | |
For male | 0 | 2 | 0 | 0 | 4 | 6 | 2 | 3 | 6 | 4 | 6 |
Price per unit of goods | 3 | 5 | 3 | 2 | 5 | 5 | 3 | 3 | 5 | 6 | 4 |
Total value of goods | 0 | 10 | 0 | 0 | 20 | 30 | 6 | 9 | 30 | 24 | 24 |
TABLE 5
As shown in table 7, a total value data table of each commodity for the category of the female user according to the embodiment of the present invention;
love | Suspense questions of the present invention | Art and literature | Comprehensive art | Swordsman | Movement of | History of | Youth of youth | Crime | War | Cartoon | |
Woman | 6 | 1 | 4 | 18 | 2 | 2 | 1 | 10 | 2 | 0 | 2 |
Price per unit of goods | 3 | 5 | 3 | 2 | 5 | 5 | 3 | 3 | 5 | 6 | 4 |
Total value of goods | 18 | 5 | 12 | 36 | 10 | 10 | 3 | 30 | 10 | 6 | 8 |
TABLE 6
Step4, calculating the association weight, frequency weight and value weight of each commodity of each user category;
wherein ,(W1(cx))iRepresenting a user class cxPurchased article GiI.e. user class cxFor commodity GiAssociated purchase frequency (R2 (c)x))iIn the user class cxFor commodity GiTotal frequency of purchases (R3 (c)x))iThe ratio of (A) to (B);
in the embodiment of the invention, the rootAccording to tables 3 and 4, the associated weight of the commodity "love" purchased by the female user is
wherein ,(V1(cx))iRepresenting a user class cxPurchased article GiFrequency weight of (2), i.e. user class cxFor commodity GiThe total purchase frequency of (R3 (c)x))iIn the user class cxTotal frequency of purchases for all itemsThe ratio of (A) to (B);
in the embodiment of the present invention, according to table 3, the frequency weight of the "love" of the commodity purchased by the female user is
(U1(cx))iRepresenting a user class cxPurchased article GiValue weight of (1), i.e. user class cxPurchased article GiTotal value of (Val3 (c)x))iIn the user class cxTotal value of all purchased goodsThe ratio of (A) to (B);
in the embodiment of the present invention, according to table 6, the value weight of the "love" of the commodity purchased by the female user is
Step5, counting the frequency of the simultaneous purchasing of the commodities of each user type according to the commodity transaction data of the user; and creates a directed commodity purchase frequency matrix GG1 (c) for each user categoryx);
Extracting user category c from user commodity transaction dataxBased on the user category cxCreating a user category cxThe directional commodity purchase frequency matrix;
commodity GiAnd GjN x n matrix GG1 (c) for determining whether there is a directed simultaneous purchase trigger relationship betweenx)={(R4(cx))ijRepresents; (R4 (c)x))ijThe value of (D) represents the user class cxPurchase of goods G in transaction data ofiWhether or not to trigger the purchase of the commodity Gj;(R4(cx))ij1 denotes purchase of an article GiTime triggered purchase of merchandise Gj;(R4(cx))ij0 denotes the purchase of the product GiWithout triggering the purchase of the commodity Gj;i={1,2,3...n},j={1,2,3...n};
In the embodiment of the invention, in the transaction data of which the user category is female, a commodity purchasing frequency matrix is shown in a table 7;
love | Suspense questions of the present invention | Art and literature | Comprehensive art | Swordsman | Movement of | History of | Youth of youth | Crime | War | Cartoon | |
Love | / | 1 | 1 | 2 | |||||||
Suspense questions of the present invention | / | ||||||||||
Art and literature | / | 1 | |||||||||
Comprehensive art | 1 | 2 | / | 3 | 1 | ||||||
Swordsman | / | 1 | |||||||||
Movement of | / | ||||||||||
History of | / | ||||||||||
Youth of youth | 1 | / | |||||||||
Crime | 1 | / | |||||||||
War | / | ||||||||||
Cartoon | / |
TABLE 7
Step6, according to the directional commodity purchasing frequency moment of each user categoryArray GG1 (c)x) Creating a undirected merchandise purchase frequency matrix GG2 for each user category (c)x);
User class cxThe undirected commodity purchase frequency matrix of (2) is represented by an n × n matrix GG2 (c)x)={(R5(cx))ijRepresents; (R5 (c)x))ijThe value of (D) represents the user class cxIn transaction data of (1) commodity GiAnd GjFrequency of concurrent purchases;
commodity GiAnd GjThe presence of a simultaneous purchase trigger relationship includes: purchasing goods GiTime triggered purchase of merchandise GjAnd purchase of goods GjTime triggered purchase of merchandise GiThe relationship of (1);
in the embodiment of the invention, in the transaction data of which the user category is female, the purchase frequency matrix of the undivided commodities is shown in a table 8;
love | Suspense questions of the present invention | Art and literature | Comprehensive art | Swordsman | Movement of | History of | Youth of youth | Crime | War | Cartoon | |
Love | / | 1 | 2 | 2 | |||||||
Suspense questions of the present invention | / | ||||||||||
Art and literature | 1 | / | 3 | ||||||||
Comprehensive art | 2 | 3 | / | 4 | 1 | ||||||
Swordsman | / | 1 | 1 | ||||||||
Movement of | 1 | / | |||||||||
History of | / | ||||||||||
Youth of youth | 2 | 4 | / | ||||||||
Crime | 1 | / | |||||||||
War | / | ||||||||||
Cartoon | 1 | / |
TABLE 8
Step7, directed commodity purchase frequency matrix GG1 for each user category (c)x) Normalization processing is carried out to obtain a corresponding directional commodity purchase weighting frequency matrix GG11 (c)x);
For user class cxDirected commodity purchase frequency matrix GG1 (c)x) The normalized calculation formula is as follows:
wherein ,(F2x)ijRepresenting a user class cxPurchasing goods GiTime triggered purchase of merchandise GjThe normalized frequency order value of (1) is also called a weighted frequency order value; (R4 (c)x))ijRepresenting a user class cxPurchasing goods GiTime triggered purchase of merchandise GjThe frequency of (2); min (GG1 (c)x) Indicating a directional commodity purchase frequency matrix GG1 (c)x) Minimum order of frequencies; max (GG1 (c)x) A matrix GG1 (c) indicating the frequency of purchasing commoditiesx) Maximum frequency order value of;
in the embodiment of the present invention, the directional commodity purchase weighting frequency matrix data in which the user category is girl is shown in table 9;
love | Suspense questions of the present invention | Art and literature | Comprehensive art | Swordsman | Movement of | History of | Youth of youth | Crime | War | Cartoon | |
Love | / | 0.1 | 0.1 | 0.2 | |||||||
Suspense questions of the present invention | / | ||||||||||
Art and literature | / | 0.1 | |||||||||
Comprehensive art | 0.1 | 0.2 | / | 0.3 | 0.1 | ||||||
Swordsman | / | 0.1 | |||||||||
Movement of | / | ||||||||||
History of | / | ||||||||||
Youth of youth | 0.1 | / | |||||||||
Crime | 0.1 | / | |||||||||
War | / | ||||||||||
Cartoon | / |
Table 9Step8, calculates the trigger center degree (D) of each commodity for each user category1(cx))i;
wherein ,(D1(cx))iRepresenting a user class cxPurchased article GiTriggering centrality of (3); (CF (c)x))iRepresenting a user class cxPurchased article GiDegree of triggering, i.e. user class cxPurchasing goods GiThe normalized frequency of purchasing other commodity types is triggered; wherein i≠j;
step9, calculating the network core degree (H) of each commodity of each user category1(cx))i;
Fig. 2 is a flowchart of a method for calculating a network core degree of each commodity of each user category according to an embodiment of the present invention, including the following steps:
s91, GG22 (c) according to the undivided commodity purchase frequency matrixx) Building a commodity connection relation network;
the method specifically comprises the following steps: establishing a commodity connection relation network by taking commodities as nodes, taking the connection relation among the commodities as edges and the frequency of simultaneous purchase among the commodities as edge weights;
if commodity node Gi and GjThere is a simultaneous purchase triggering relationship (direct connection relationship) between them, the commodity node Gi and GjA connecting edge is added between the two edges; commodity node Gi and GjThe intermediate nodes are connected by a path which comprises at least 1 intermediate node and at least 2 connecting edges, and the number of the paths is at least one;
s92, for the commodity node G with direct or indirect connectioni and GjObtaining a commodity node GjRelative node GiNode level lev;
let GiIs the original node, if the commodity node Gi and GjThere is a direct connection relationship between them, node GjIs GiA level 1 node of (1);
if commodity node Gi and GjThere is indirect connection relation and connect through m node paths, m is greater than or equal to 1, P ═ path for m node path set1,path2,….GmRepresents; the number of connection edges included in each node path in P is S ═ side1,side2,….sidemDenotes, then node GjRelative node GiThe node level of (c) lev ═ min(s), min(s) is not less than 2, and min () represents the minimum value;
FIG. 3 shows a commodity node G provided in the embodiment of the present invention1、G2、G3、G4、G5A connection relationship network diagram of (1);
wherein G1 and G2If there is a direct connection, G2Is G1A level 1 node of (1);
G1 and G3An indirect connection relation exists, and the first node path and the second node path can be indirectly connected; the first node path includes 2 connecting edges: g1-G2 and G2-G3(ii) a The second node path includes 4 connecting edges: g1-G2、G2-G4、G4-G5 and G5-G3(ii) a Then G is3Is G1A level 2 node of (1);
the value on the connecting edge represents the edge weight (i.e., the frequency of purchases at the same time) between two directly connected commodity nodes, such as G1 and G2The edge weight of (1) is 10, G2 and G3The edge weight of (2) is 5;
s93, calculating the network median degree F of each commodity node of each user categoryb(i);
Wherein, (FB (c)x))iRepresenting a user class cxCommodity node GiThe median level in the network of (2); levijPresentation and merchandise node GiDirect or indirect connected commodity node GjNode level of (2); (R3 (c)x))jPresentation and merchandise node GiDirect or indirect connected commodity node GjTotal purchase frequency of corresponding goods, i.e. user class cxFor commodity GjTotal frequency of purchases; n represents the total number of commodity nodes;
s94, calculating network core degree (H) of each commodity node of each user type1(cx))i;
wherein ,(H1(cx))iRepresenting a user class cxCommodity node GiThe network core degree of (c);
step10, associating the weight (W) of each commodity with each user category1(cx))iFrequency weight (V)1(cx))iValue weight (U)1(cx))iTrigger centrality (D)1(cx))iNetwork core degree (H)1(cx))iNormalization processing is carried out to obtain corresponding weighted associated weight (W)2(cx))iWeighted frequency weight (V)2(cx))iWeighted value weight (U)2(cx))iWeighted trigger centrality (D)2(cx))iWeighted network core degree (H)2(cx))i;
wherein, F represents the value after normalization processing; r respectively represents a user category cxPurchasing goods GiAssociated weight (W)1(cx))iFrequency weight (V)1(cx))iValue weight (U)1(cx))iTrigger centrality (D)1(cx))iNetwork core degree (H)1(cx))i(ii) a min (T) respectively represents the minimum value in the association weight, the frequency weight, the value weight, the triggering centrality and the network core degree data of all the commodities; max (T) respectively represents the maximum numerical value in the data of the association weight, the frequency weight, the value weight, the triggering centrality and the network core degree of all the commodities;
for example, if F represents normalized network core, then R represents user class cxCommodity node GiNetwork core ofDegree (H)1(cx))i(ii) a min represents user class cx(ii) a network core degree set of all commodity nodes { (H)1(cx))iI | i | 1,2,3 … … n } and max represents the user category cx(ii) a network core degree set of all commodity nodes { (H)1(cx))iI | i ═ 1,2,3 … … n } maximum; the rest is analogized in the same way;
step11, weight associating weight (W)2(cx))iWeighted frequency weight (V)2(cx))iWeighted value weight (U)2(cx))iWeighted trigger centrality (D)2(cx))iWeighted network core degree (H)2(cx))iWeighted summation is carried out to obtain the optimal node degree (S (c)) for classifying the user commoditiesx))i;
(S(cx))i=((W2(cx))i+(V2(cx))i+(U2(cx))i+(D2(cx))i+(H2(cx))i)/5;
Wherein (S (c)x))iRepresenting a user class cxCommodity node GiThe optimal node degree of (2);
alternatively,
calculating a weighted relevance weight (W) from an analytic hierarchy process2(cx))iWeighted frequency weight (V)2(cx))iWeighted value weight (U)2(cx))iWeighted trigger centrality (D)2(cx))iWeighted network core degree (H)2(cx))iWeights in the goods recommendation decision target are weighted and associated with weights (W)2(cx))iWeighted frequency weight (V)2(cx))iWeighted value weight (U)2(cx))iWeighted trigger centrality (D)2(cx))iWeighted network core degree (H)2(cx))iWeighted summation is carried out to obtain the optimal node degree (S (c)) for classifying the user commoditiesx))i;
(S(cx))i=α1(W2(cx))i+α2(V2(cx))i+α3(U2(cx))i+α4(D2(cx))i+α5(H2(cx))i;
wherein ,α1、α2、α3、α4、α5Respectively represent (W)2(cx))i、(V2(cx))i、(U2(cx))i、(D2(cx))i、(H2(cx))iWeights in the goods recommendation decision target;
specifically, the commodity recommendation is used as a decision target, and the weight (W) is weighted and associated2(cx))iWeighted frequency weight (V)2(cx))iWeighted value weight (U)2(cx))iWeighted trigger centrality (D)2(cx))iWeighted network core degree (H)2(cx))iThe evaluation matrix is used as a judgment matrix element in an analytic hierarchy process, and the importance degree of the element is subjected to pairwise comparison evaluation; the analytic hierarchy process is a conventional process, and is described herein again;
specifically, in an application scenario, the judgment is performed according to specific services, and specific evaluation filling of the judgment model is shown in table 10;
watch 10
Weighted relevance weights (W) calculated from analytic hierarchy process2(cx))iWeighted frequency weight (V)2(cx))iWeighted value weight (U)2(cx))iWeighted trigger centrality (D)2(cx))iWeighted network core degree (H)2(cx))iWeight α of1、α2、α3、α4、α5Respectively as follows: 0.55, 023, 0.08, 0.08, 0.08;
the best node degree of the user' S goods is classified (S (c)x))iComprises the following steps:
(S(cx))i=0.55(W2(cx))i+0.23(V2(cx))i+0.08(U2(cx))i+0.08(D2(cx))i+0.08(H2(cx))i;
and Step12, recommending the commodities to the corresponding classified users according to the optimal node degree of the commodities of the classified users.
Specifically, in the embodiment, the commodities are sorted according to the size of the optimal node degree of the commodities, and the commodities with the large optimal node degree are selected to be preferentially pushed to corresponding classified users;
corresponding to the above-mentioned commodity recommendation method based on user attributes and commodity types, fig. 4 is a structure diagram of a commodity recommendation system based on user attributes and commodity types according to an embodiment of the present invention; the system comprises:
the user commodity associated information table creating module is used for creating a classified user commodity associated information table according to the user commodity transaction data;
the commodity purchase frequency counting module is used for counting the total purchase frequency of each commodity for each user category; counting the associated purchase frequency of each user category to each commodity;
the commodity total value calculating module is used for calculating the total value of each commodity according to the total purchase frequency of each user type to each commodity and the unit price of each commodity;
the weight calculation module is used for calculating the association weight, the frequency weight and the value weight of each commodity of each user category;
the oriented commodity purchase frequency matrix creating module is used for counting the oriented and simultaneous purchase frequency of commodities of various user categories according to the commodity transaction data of the users; creating a directional commodity purchasing frequency matrix of each user category;
the undirected commodity purchasing frequency matrix creating module is used for creating an undirected commodity purchasing frequency matrix of each user category according to the directed commodity purchasing frequency matrix of each user category;
the first normalization processing module is used for performing normalization processing on the directional commodity purchasing frequency matrixes of all user categories to obtain corresponding directional commodity purchasing weighting frequency matrixes;
the triggering centrality calculating module is used for calculating the triggering centrality of each commodity of each user category;
a network core degree calculation device for calculating the network core degree of each commodity of each user category;
the second normalization processing module is used for respectively normalizing the association weight, the frequency weight, the value weight, the triggering centrality and the network core degree of each commodity of each user category to obtain the corresponding weighted association weight, the weighted frequency weight, the weighted value weight, the weighted triggering centrality and the weighted network core degree;
the optimal node degree calculation module is used for carrying out weighted summation on the weighted association weight, the weighted frequency weight, the weighted value weight, the weighted triggering centrality and the weighted network core degree to obtain the optimal node degree of the classified user commodities;
and the commodity recommending module is used for recommending commodities to the corresponding classified users according to the optimal node degree of the commodities of the classified users.
Further, the user commodity transaction data includes fields: user number, user category, commodity category and transaction time;
the classification user commodity association information table comprises the following fields: a user number, a user category and a plurality of specific commodity categories; when the value in the user commodity association information table indicates that one-time purchasing behavior occurs, whether the corresponding row user purchases the corresponding commodity or not is judged, 1 indicates purchasing, and 0 indicates not purchasing; each row in the table represents a purchase record; the same user can have multiple purchase records;
further, in the embodiment of the present invention, it is assumed that the commodity platform has n kinds of commodities in total, and the set G ═ G is used1,G2,….GnRepresents; n represents the number of commodity types, also called the number of commodity nodes; there are a total of k user categories, with the set C ═ C1,c2,….ckDenotes, i ═ 1,2, 3.. n }, and x ═ 1,2, 3.. k };
user class cxFor commodity GiIs given by k × n matrix CF3 { (R3 (c)x))iDenotes (R3 (c)x))iThe value of (D) represents the user class cxFor commodity GiTotal frequency of purchases;
user class cxFor commodity GiIs given by k × n matrix CF2 { (R2 (c)x))iDenotes (R2 (c)x))iThe value of (D) represents the user class cxFor commodity GiAssociated purchase frequency of;
(R2(cx))i=(R3(cx))i-(R1(cx))i(ii) a Wherein (R1 (c)x))iRepresenting a user class cxFor commodity GiIndividual purchase frequency of; commodity GiThe individual purchase frequency of (A) indicates that only a single commodity G is purchasediThe frequency of (2); commodity GiThe associated purchase frequency of (2) indicates that the purchased product includes product GiFrequency of (i.e. article G)iFrequency of purchases at the same time with all other merchandise;
further, user category cxPurchased article GiThe total value of (a) is determined by k × n matrix CF4 { (Val3 (c)x))iDenotes (Val3 (c)x))iThe value of (D) represents the user class cxPurchased article GiThe total value of (c); (Val3 (c)x))i=(R3(cx))i×(Pr)i;
Further, the calculating of the association weight, the frequency weight, and the value weight of each commodity of each user category specifically includes:
wherein ,(W1(cx))iRepresenting a user class cxPurchased article GiI.e. user class cxFor commodity GiAssociated purchase frequency (R2 (c)x))iIn the user class cxFor commodity GiTotal frequency of purchases (R3 (c)x))iThe ratio of (A) to (B);
wherein ,(V1(cx))iRepresenting a user class cxPurchased article GiFrequency weight of (2), i.e. user class cxFor commodity GiThe total purchase frequency of (R3 (c)x))iIn the user class cxTotal frequency of purchases for all itemsThe ratio of (A) to (B);
(U1(cx))irepresenting a user class cxPurchased article GiValue weight of (1), i.e. user class cxPurchased article GiTotal value of (Val3 (c)x))iIn the user class cxTotal value of all purchased goodsThe ratio of (A) to (B);
further, according to the user commodity transaction data, counting the frequency of the commodities of each user category which are purchased simultaneously; and creates a directed commodity purchase frequency matrix GG1 (c) for each user categoryx) The method specifically comprises the following steps:
extracting user category c from user commodity transaction dataxBased on the user category cxCreating a user category cxThe directional commodity purchase frequency matrix;
commodity GiAnd GjN x n matrix GG1 (c) for determining whether there is a directed simultaneous purchase trigger relationship betweenx)={(R4(cx))ijRepresents; (R4 (c)x))ijThe value of (D) represents the user class cxPurchase of goods G in transaction data ofiWhether or not to trigger the purchase of the commodity Gj;(R4(cx))ij1 denotes purchase of an article GiTime triggered purchase of merchandise Gj;(R4(cx))ij0 denotes the purchase of the product GiWithout triggering the purchase of the commodity Gj;i={1,2,3...n},j={1,2,3...n};
Further, the directional commodity purchase frequency matrix GG1 (c) according to the user categoriesx) Creating a undirected merchandise purchase frequency matrix GG2 for each user category (c)x) The method specifically comprises the following steps:
user class cxThe undirected commodity purchase frequency matrix of (2) is represented by an n × n matrix GG2 (c)x)={(R5(cx))ijRepresents; (R5 (c)x))ijThe value of (D) represents the user class cxIn transaction data of (1) commodity GiAnd GjFrequency of concurrent purchases;
commodity GiAnd GjThe presence of a simultaneous purchase trigger relationship includes: purchasing goods GiTime triggered purchase of merchandise GjAnd purchase of goods GjTime triggered purchase of merchandise GiThe relationship of (1);
further, the directional commodity purchase frequency matrix GG1 for each user category (c)x) Normalization processing is carried out to obtain a corresponding directional commodity purchase weighting frequency matrix GG11 (c)x) The method specifically comprises the following steps:
for user class cxDirected commodity purchase frequency matrix GG1 (c)x) The normalized calculation formula is as follows:
wherein ,(F2x)ijRepresenting a user class cxPurchasing goods GiTime triggered purchase of merchandise GjThe normalized frequency order value of (1) is also called a weighted frequency order value; (R4 (c)x))ijRepresenting a user class cxPurchasing goods GiTime triggered purchase of merchandise GjThe frequency of (2); min (GG1 (c)x) Indicating a directional commodity purchase frequency matrix GG1 (c)x) Minimum order of frequencies; max (GG1 (c)x) A matrix GG1 (c) indicating the frequency of purchasing commoditiesx) Maximum frequency order value of;
further, the triggering centrality (D) of each commodity of each user category is calculated1(cx))iThe method specifically comprises the following steps:
wherein ,(D1(cx))iRepresenting a user class cxPurchased article GiTriggering centrality of (3); (CF (c)x))iRepresenting a user class cxPurchased article GiDegree of triggering, i.e. user class cxPurchasing goods GiThe normalized frequency of purchasing other commodity types is triggered; wherein i≠j;
further, fig. 5 is a structural diagram of a network core degree computing device according to an embodiment of the present invention; the network core degree calculating device comprises:
a commodity connection relation network construction module used for GG22 (c) according to the undivided commodity purchasing frequency matrixx) Building a commodity connection relation network;
the method specifically comprises the following steps: establishing a commodity connection relation network by taking commodities as nodes, taking the connection relation among the commodities as edges and the frequency of simultaneous purchase among the commodities as edge weights;
if commodity node Gi and GjThere is a simultaneous purchase triggering relationship (direct connection relationship) between them, the commodity node Gi and GjA connecting edge is added between the two edges; commodity node Gi and GjThe intermediate nodes are connected by a path which comprises at least 1 intermediate node and at least 2 connecting edges, and the number of the paths is at least one;
a node grade acquisition module for the commodity node G with direct or indirect connectioni and GjObtaining a commodity node GjRelative node GiNode level lev;
let GiIs the original node, if the commodity node Gi and GjThere is a direct connection relationship between them, node GjIs GiA level 1 node of (1);
if commodity node Gi and GjThere is indirect connection relation and connect through m node paths, m is greater than or equal to 1, P ═ path for m node path set1,path2,….GmRepresents; the number of connection edges included in each node path in P is S ═ side1,side2,….sidemDenotes, then node GjRelative node GiThe node level of (c) lev ═ min(s), min(s) is not less than 2, and min () represents the minimum value;
a network median calculation module for calculating network median F of each commodity node of each user categoryb(i);
Wherein, (FB (c)x))iRepresenting a user class cxCommodity node GiThe median level in the network of (2); levijPresentation and merchandise node GiDirect or indirect connected commodity node GjNode level of (2); (R3 (c)x))jPresentation and merchandise node GiDirect or indirect connected commodity node GjTotal purchase frequency of corresponding goods, i.e. user class cxFor commodity GjTotal frequency of purchases; n represents the total number of commodity nodes;
a network core degree calculation module for calculating the network core degree (H) of each commodity node of each user category1(cx))i;
Further, the user categories are associated with the commodities by weight (W)1(cx))iFrequency weight (V)1(cx))iValue weight (U)1(cx))iTrigger centrality (D)1(cx))iNetwork core degree (H)1(cx))iNormalization processing is carried out to obtain corresponding weighted associated weight (W)2(cx))iWeighted frequency weight (V)2(cx))iWeighted value weight (U)2(cx))iWeighted trigger centrality (D)2(cx))iWeighted network core degree (H)2(cx))i;
wherein, F represents the value after normalization processing; r respectively represents a user category cxPurchasing goods GiAssociated weight (W)1(cx))iFrequency weight (V)1(cx))iValue weight (U)1(cx))iTrigger centrality (D)1(cx))iNetwork core degree (H)1(cx))i(ii) a min (T) represents the associated weight, frequency weight, value weight, triggering centrality and network core degree of all the commoditiesA minimum value of (d); max (T) respectively represents the maximum numerical value in the data of the association weight, the frequency weight, the value weight, the triggering centrality and the network core degree of all the commodities;
for example, if F represents normalized network core, then R represents user class cxCommodity node GiNetwork core degree (H)1(cx))i(ii) a min represents user class cx(ii) a network core degree set of all commodity nodes { (H)1(cx))iI | i | 1,2,3 … … n } and max represents the user category cx(ii) a network core degree set of all commodity nodes { (H)1(cx))iI | i ═ 1,2,3 … … n } maximum; the rest is analogized in the same way;
further, the weight (W) is weighted and associated2(cx))iWeighted frequency weight (V)2(cx))iWeighted value weight (U)2(cx))iWeighted trigger centrality (D)2(cx))iWeighted network core degree (H)2(cx))iWeighted summation is carried out to obtain the optimal node degree (S (c)) for classifying the user commoditiesx))i(ii) a The method specifically comprises the following steps:
(S(cx))i=((W2(cx))i+(V2(cx))i+(U2(cx))i+(D2(cx))i+(H2(cx))i)/5;
further, the "will weight the associated weight (W)2(cx))iWeighted frequency weight (V)2(cx))iWeighted value weight (U)2(cx))iWeighted trigger centrality (D)2(cx))iWeighted network core degree (H)2(cx))iWeighted summation is carried out to obtain the optimal node degree (S (c)) for classifying the user commoditiesx))i"can be replaced by:
"computing weighted correlations according to analytic hierarchy ProcessWeight (W)2(cx))iWeighted frequency weight (V)2(cx))iWeighted value weight (U)2(cx))iWeighted trigger centrality (D)2(cx))iWeighted network core degree (H)2(cx))iWeights in the goods recommendation decision target are weighted and associated with weights (W)2(cx))iWeighted frequency weight (V)2(cx))iWeighted value weight (U)2(cx))iWeighted trigger centrality (D)2(cx))iWeighted network core degree (H)2(cx))iWeighted summation is carried out to obtain the optimal node degree (S (c)) for classifying the user commoditiesx))i;”;
(S(cx))i=α1(W2(cx))i+α2(V2(cx))i+α3(U2(cx))i+α4(D2(cx))i+α5(H2(cx))i wherein ,α1、α2、α3、α4、α5Respectively represent (W)2(cx))i、(V2(cx))i、(U2(cx))i、(D2(cx))i、(H2(cx))iWeights in the goods recommendation decision target;
specifically, the commodity recommendation is used as a decision target, and the weight (W) is weighted and associated2(cx))iWeighted frequency weight (V)2(cx))iWeighted value weight (U)2(cx))iWeighted trigger centrality (D)2(cx))iWeighted network core degree (H)2(cx))iThe evaluation matrix is used as a judgment matrix element in an analytic hierarchy process, and the importance degree of the element is subjected to pairwise comparison evaluation; the analytic hierarchy process is a conventional process, and is described herein again;
specifically, in an application scenario, the judgment is performed according to specific services, and specific evaluation filling of the judgment model is shown in table 10;
watch 10
Weighted relevance weights (W) calculated from analytic hierarchy process2(cx))iWeighted frequency weight (V)2(cx))iWeighted value weight (U)2(cx))iWeighted trigger centrality (D)2(cx))iWeighted network core degree (H)2(cx))iWeight α of1、α2、α3、α4、α5Respectively as follows: 0.55, 023, 0.08, 0.08, 0.08;
the best node degree of the user' S goods is classified (S (c)x))iComprises the following steps:
(S(cx))i=0.55(W2(cx))i+0.23(V2(cx))i+0.08(U2(cx))i+0.08(D2(cx))i+0.08(H2(cx))i;
further, the commodity recommendation is performed on the corresponding classified users according to the optimal node degrees of the commodities of the classified users, specifically in the embodiment, the commodities are sorted according to the optimal node degrees of the commodities, and the commodities with the high optimal node degrees are selected to be preferentially pushed to the corresponding classified users;
an embodiment of the present invention further provides a terminal device, where the terminal device includes: a processor, a memory, and a computer program stored in the memory and executable on the processor. The processor, when executing the computer program, implements the steps in one embodiment of the merchandise recommendation method based on the user attribute and the merchandise type, such as steps 1 to Step12 shown in fig. 1. Or, the processor, when executing the computer program, implements the functions of the modules in the above-mentioned product recommendation system embodiment based on the user attribute and the product type.
It will be understood by those skilled in the art that all or part of the steps in the method according to the above embodiments may be implemented by hardware related to program instructions, and the program may be stored in a computer readable storage medium, such as ROM, RAM, magnetic disk, optical disk, etc.
The sequence number of each step in the foregoing embodiments does not mean the execution sequence, and the execution sequence of each process should be determined by the function and the internal logic of the process, and should not constitute any limitation on the implementation process of the embodiments of the present invention.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents and improvements made within the spirit and principle of the present invention are intended to be included within the scope of the present invention.
Claims (16)
1. A commodity recommendation method based on user attributes and commodity types is characterized by comprising the following steps:
establishing a classified user commodity association information table according to the user commodity transaction data;
counting the total purchase frequency of each user category to each commodity; counting the associated purchase frequency of each user category to each commodity;
calculating the total value of each commodity according to the total purchase frequency of each commodity and the unit price of each commodity of each user category;
calculating the association weight, frequency weight and value weight of each commodity of each user category;
counting the frequency of the commodities of each user category which are purchased simultaneously according to the commodity transaction data of the users; creating a directional commodity purchasing frequency matrix of each user category;
creating an undirected commodity purchasing frequency matrix of each user category according to the directed commodity purchasing frequency matrix of each user category;
carrying out normalization processing on the directional commodity purchasing frequency matrixes of all user categories to obtain corresponding directional commodity purchasing weighting frequency matrixes;
calculating the triggering centrality of each commodity of each user category;
calculating the network core degree of each commodity of each user category;
respectively carrying out normalization processing on the association weight, the frequency weight, the value weight, the triggering centrality and the network core degree of each commodity of each user category to obtain the corresponding weighted association weight, weighted frequency weight, weighted value weight, weighted triggering centrality and weighted network core degree;
weighting and summing the weighted association weight, the weighted frequency weight, the weighted value weight, the weighted triggering centrality and the weighted network core degree to obtain the optimal node degree for classifying the user commodities;
and recommending the commodities to the corresponding classified users according to the optimal node degree of the commodities of the classified users.
2. The commodity recommendation method based on the user attributes and the commodity types according to claim 1, wherein the association weight, the frequency weight, the value weight, the triggering centrality and the network centrality of each commodity of each user category are normalized to obtain the corresponding weighted association weight, the weighted frequency weight, the weighted value weight, the weighted triggering centrality and the weighted network centrality, instead of:
and calculating the weights of the weighted association weight, the weighted frequency weight, the weighted value weight, the weighted triggering centrality and the weighted network core degree in the commodity recommendation decision target according to an analytic hierarchy process, and performing weighted summation on the weighted association weight, the weighted frequency weight, the weighted value weight, the weighted triggering centrality and the weighted network core degree to obtain the optimal node degree for classifying the user commodities.
3. The commodity recommendation method based on user attributes and commodity types according to claim 2, wherein said user commodity transaction data comprises fields of: user number, user category, commodity category and transaction time; the classification user commodity association information table comprises the following fields: a user number, a user category, and a number of specific merchandise categories.
4. The commodity recommendation method based on the user attributes and the commodity types according to claim 3, wherein the total purchase frequency of each commodity by each user category is counted; the method for counting the associated purchase frequency of each user category on each commodity specifically comprises the following steps:
let the commodity platform total n kinds of commodities, use set G ═ G1,G2,….GnRepresents; n represents the number of commodity types, also called the number of commodity nodes; there are a total of k user categories, with the set C ═ C1,c2,….ckDenotes, i ═ 1,2, 3.. n }, and x ═ 1,2, 3.. k };
user class cxFor commodity GiIs given by k × n matrix CF3 { (R3 (c)x))iDenotes (R3 (c)x))iThe value of (D) represents the user class cxFor commodity GiTotal frequency of purchases;
user class cxFor commodity GiIs given by k × n matrix CF2 { (R2 (c)x))iDenotes (R2 (c)x))iThe value of (D) represents the user class cxFor commodity GiAssociated purchase frequency of; commodity GiThe associated purchase frequency of (2) indicates that the purchased product includes product GiFrequency of (i.e. article G)iFrequency of simultaneous purchases with all other merchandise.
5. The item recommendation method based on the user attributes and the item types according to claim 4, wherein the calculating the total value of each item according to the total purchase frequency and the unit price of each item for each user category specifically comprises:
user class cxPurchased article GiThe total value of (a) is determined by k × n matrix CF4 { (Val3 (c)x))iDenotes (Val3 (c)x))iThe value of (D) represents the user class cxPurchased article GiThe total value of (c);
(Val3(cx))i=(R3(cx))i×(Pr)i。
6. the method for recommending commodities based on user attributes and commodity types according to claim 5, wherein the calculating of the association weight, the frequency weight and the value weight of each commodity of each user category specifically comprises:
wherein ,(W1(cx))iRepresenting a user class cxPurchased article GiI.e. user class cxFor commodity GiAssociated purchase frequency (R2 (c)x))iIn the user class cxFor commodity GiTotal frequency of purchases (R3 (c)x))iThe ratio of (A) to (B);
wherein ,(V1(cx))iRepresenting a user class cxPurchased article GiFrequency weight of (2), i.e. user class cxFor commodity GiThe total purchase frequency of (R3 (c)x))iIn the user class cxTotal frequency of purchases for all itemsThe ratio of (A) to (B);
7. The commodity recommendation method based on the user attributes and the commodity types according to claim 6, wherein the normalization processing of the directional commodity purchase frequency matrix of each user category to obtain the corresponding directional commodity purchase weighting frequency matrix specifically comprises:
for user class cxDirected commodity purchase frequency matrix GG1 (c)x) The normalized calculation formula is as follows:
wherein ,(F2x)ijRepresenting a user class cxPurchasing goods GiTime triggered purchase of merchandise GjThe normalized frequency order value of (1) is also called a weighted frequency order value; (R4 (c)x))ijRepresenting a user class cxPurchasing goods GiTime triggered purchase of merchandise GjThe frequency of (2); min (GG1 (c)x) Indicating a directional commodity purchase frequency matrix GG1 (c)x) Minimum order of frequencies; max (GG1 (c)x) A matrix GG1 (c) indicating the frequency of purchasing commoditiesx) Of the maximum frequency order.
8. The method for recommending commodities based on user attributes and commodity types according to claim 7, wherein said calculating the triggering centrality of each commodity of each user category specifically comprises:
wherein ,(D1(cx))iRepresenting a user class cxPurchased article GiTriggering centrality of (3); (CF (c)x))iRepresenting a user class cxPurchased article GiDegree of triggering, i.e. user class cxPurchasing goods GiThe normalized frequency of purchasing other commodity types is triggered; wherein i≠j.
9. The method of claim 8, wherein the calculating the network core degree of each commodity for each user category comprises:
constructing a commodity connection relation network according to the undirected commodity purchasing frequency matrix;
for commodity nodes G with direct or indirect connectionsi and GjObtaining a commodity node GjRelative node GiNode level lev;
calculating the network median degree of each commodity node of each user category;
and calculating the network core degree of each commodity node of each user class.
10. The item recommendation method based on the user attributes and the item types according to claim 9,
the purchase frequency matrix GG22 (c) according to the undivided commoditiesx) Building a commodity connection relation network, specifically:
establishing a commodity connection relation network by taking commodities as nodes, taking the connection relation among the commodities as edges and the frequency of simultaneous purchase among the commodities as edge weights;
if commodity node Gi and GjThere is a simultaneous purchase triggering relationship between them, then the commodity node Gi and GjA connecting edge is added between the two edges; commodity node Gi and GjThe intermediate nodes are connected by a path which comprises at least 1 intermediate node and at least 2 connecting edges, and the number of the paths is at least one;
for commodity nodes G with direct or indirect connectionsi and GjObtaining a commodity node GjRelative node GiNode level lev;
let GiIs the original node, if the commodity node Gi and GjThere is a direct connection relationship between them, node GjIs GiA level 1 node of (1);
if commodity node Gi and GjThere is indirect connection relation and connect through m node paths, m is greater than or equal to 1, P ═ path for m node path set1,path2,….GmRepresents; the number of connection edges included in each node path in P is S ═ side1,side2,….sidemDenotes, then node GjRelative node GiThe node level of (c) lev ═ min(s), min(s) is not less than 2, and min () represents the minimum value;
the calculation of the network medium level of each commodity node of each user category is specifically as follows:
wherein, (FB (c)x))iRepresenting a user class cxCommodity node GiThe median level in the network of (2); levijPresentation and merchandise node GiDirect or indirect connected commodity node GjNode level of (2); (R3 (c)x))jPresentation and merchandise node GiDirect or indirect connected commodity node GjTotal purchase frequency of corresponding goods, i.e. user class cxFor commodity GjTotal frequency of purchases; n represents the total number of commodity nodes;
the method for calculating the network core degree of each commodity node of each user category specifically comprises the following steps:
wherein ,(H1(cx))iRepresenting a user class cxCommodity node GiNetwork core degree of (2).
11. The commodity recommendation method based on user attributes and commodity types according to claim 10, wherein the association weight, the frequency weight, the value weight, the triggering centrality, and the network core degree of each commodity of each user category are normalized respectively,
wherein, F represents the value after normalization processing; r respectively represents a user category cxPurchasing goods GiAssociated weight (W)1(cx))iFrequency weight (V)1(cx))iValue weight (U)1(cx))iTrigger centrality (D)1(cx))iNetwork core degree (H)1(cx))i(ii) a min (T) respectively represents the minimum value in the association weight, the frequency weight, the value weight, the triggering centrality and the network core degree data of all the commodities; max (t) represents the maximum value among the association weight, frequency weight, value weight, trigger centrality, and network core data of all the commodities, respectively.
12. The commodity recommendation method based on user attributes and commodity types according to any one of claims 1 or 3 to 11, wherein the weighted association weight, the weighted frequency weight, the weighted value weight, the weighted trigger centrality, and the weighted network core (H) are assigned2(cx))iWeighted summation is carried out to obtain the optimal node degree (S (c)) for classifying the user commoditiesx))i;
(S(cx))i=((W2(cx))i+(V2(cx))i+(U2(cx))i+(D2(cx))i+(H2(cx))i)/5;
Wherein (S (c)x))iRepresenting a user class cxCommodity sectionPoint GiThe optimal node degree of (c).
13. The commodity recommendation method based on user attributes and commodity types according to any one of claims 1 to 11, wherein weights of the weighted association weight, the weighted frequency weight, the weighted value weight, the weighted triggering centrality and the weighted network core degree in a commodity recommendation decision target are calculated according to an analytic hierarchy process, and the weighted association weight, the weighted frequency weight, the weighted value weight, the weighted triggering centrality and the weighted network core degree are subjected to weighted summation to obtain an optimal node degree for classifying the commodities of the user;
(S(cx))i=α1(W2(cx))i+α2(V2(cx))i+α3(U2(cx))i+α4(D2(cx))i+α5(H2(cx))i wherein ,α1、α2、α3、α4、α5Respectively represent (W)2(cx))i、(V2(cx))i、(U2(cx))i、(D2(cx))i、(H2(cx))iWeight in a goods recommendation decision target.
14. A merchandise recommendation system based on user attributes and merchandise types, the system comprising:
the user commodity associated information table creating module is used for creating a classified user commodity associated information table according to the user commodity transaction data;
the commodity purchase frequency counting module is used for counting the total purchase frequency of each commodity for each user category; counting the associated purchase frequency of each user category to each commodity;
the commodity total value calculating module is used for calculating the total value of each commodity according to the total purchase frequency of each user type to each commodity and the unit price of each commodity;
the weight calculation module is used for calculating the association weight, the frequency weight and the value weight of each commodity of each user category;
the oriented commodity purchase frequency matrix creating module is used for counting the oriented and simultaneous purchase frequency of commodities of various user categories according to the commodity transaction data of the users; creating a directional commodity purchasing frequency matrix of each user category;
the undirected commodity purchasing frequency matrix creating module is used for creating an undirected commodity purchasing frequency matrix of each user category according to the directed commodity purchasing frequency matrix of each user category;
the first normalization processing module is used for performing normalization processing on the directional commodity purchasing frequency matrixes of all user categories to obtain corresponding directional commodity purchasing weighting frequency matrixes;
the triggering centrality calculating module is used for calculating the triggering centrality of each commodity of each user category;
a network core degree calculation device for calculating the network core degree of each commodity of each user category;
the second normalization processing module is used for respectively normalizing the association weight, the frequency weight, the value weight, the triggering centrality and the network core degree of each commodity of each user category to obtain the corresponding weighted association weight, the weighted frequency weight, the weighted value weight, the weighted triggering centrality and the weighted network core degree;
the optimal node degree calculation module is used for carrying out weighted summation on the weighted association weight, the weighted frequency weight, the weighted value weight, the weighted triggering centrality and the weighted network core degree to obtain the optimal node degree of the classified user commodities;
and the commodity recommending module is used for recommending commodities to the corresponding classified users according to the optimal node degree of the commodities of the classified users.
15. The commodity recommendation system according to claim 14, wherein said network core degree calculating means comprises:
the commodity connection relation network construction module is used for constructing a commodity connection relation network according to the undirected commodity purchasing frequency matrix;
a node grade acquisition module for the commodity node G with direct or indirect connectioni and GjObtaining a commodity node GjRelative node GiNode level lev;
the network median degree calculating module is used for calculating the network median degree of each commodity node of each user category;
and the network core degree calculating module is used for calculating the network core degree of each commodity node of each user type.
16. An apparatus comprising a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor when executing the computer program implements the steps of the item recommendation method based on user attributes and item types as claimed in any one of claims 1 to 13.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201911304899.5A CN110992153B (en) | 2019-12-17 | 2019-12-17 | Commodity recommendation method, system and equipment based on user attribute and commodity type |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201911304899.5A CN110992153B (en) | 2019-12-17 | 2019-12-17 | Commodity recommendation method, system and equipment based on user attribute and commodity type |
Publications (2)
Publication Number | Publication Date |
---|---|
CN110992153A true CN110992153A (en) | 2020-04-10 |
CN110992153B CN110992153B (en) | 2023-06-16 |
Family
ID=70095087
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201911304899.5A Active CN110992153B (en) | 2019-12-17 | 2019-12-17 | Commodity recommendation method, system and equipment based on user attribute and commodity type |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110992153B (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112052403A (en) * | 2020-09-11 | 2020-12-08 | 深圳市梦网视讯有限公司 | Commodity community classification-based link propagation method, system and equipment |
Citations (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2010073170A (en) * | 2008-09-22 | 2010-04-02 | Yahoo Japan Corp | Recommended commodity selection device, recommended commodity selection program and commodity retrieval device |
KR20110019289A (en) * | 2009-08-19 | 2011-02-25 | 서울대학교산학협력단 | Goods recommendation system and method considering price of goods |
JP2011070396A (en) * | 2009-09-25 | 2011-04-07 | Brother Industries Ltd | Method of analyzing trend of using commodity, method of recommending commodity, system for analyzing trend of using commodity, and system for recommending commodity |
WO2013043714A1 (en) * | 2011-09-19 | 2013-03-28 | Klever Marketing, Inc. | System and method for influencing consumer purchasing of consumer packaged goods |
WO2016052149A1 (en) * | 2014-09-29 | 2016-04-07 | 富士フイルム株式会社 | Commodity recommendation device and commodity recommendation method |
WO2016206106A1 (en) * | 2015-06-26 | 2016-12-29 | 深圳市华阳信通科技发展有限公司 | Smart apparatus and associated product recommendation method |
CN106446943A (en) * | 2016-09-19 | 2017-02-22 | 大连海事大学 | Commodity correlation big data sparse network quick clustering method |
CN107220880A (en) * | 2017-05-26 | 2017-09-29 | 杭州再顾科技有限公司 | Method, equipment and server for showing commodity in shopping at network |
CN109242633A (en) * | 2018-09-20 | 2019-01-18 | 阿里巴巴集团控股有限公司 | A kind of commodity method for pushing and device based on bigraph (bipartite graph) network |
CN110009457A (en) * | 2019-04-09 | 2019-07-12 | 昆山古鳌电子机械有限公司 | A kind of commercial product recommending system based on big data |
-
2019
- 2019-12-17 CN CN201911304899.5A patent/CN110992153B/en active Active
Patent Citations (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2010073170A (en) * | 2008-09-22 | 2010-04-02 | Yahoo Japan Corp | Recommended commodity selection device, recommended commodity selection program and commodity retrieval device |
KR20110019289A (en) * | 2009-08-19 | 2011-02-25 | 서울대학교산학협력단 | Goods recommendation system and method considering price of goods |
JP2011070396A (en) * | 2009-09-25 | 2011-04-07 | Brother Industries Ltd | Method of analyzing trend of using commodity, method of recommending commodity, system for analyzing trend of using commodity, and system for recommending commodity |
WO2013043714A1 (en) * | 2011-09-19 | 2013-03-28 | Klever Marketing, Inc. | System and method for influencing consumer purchasing of consumer packaged goods |
WO2016052149A1 (en) * | 2014-09-29 | 2016-04-07 | 富士フイルム株式会社 | Commodity recommendation device and commodity recommendation method |
WO2016206106A1 (en) * | 2015-06-26 | 2016-12-29 | 深圳市华阳信通科技发展有限公司 | Smart apparatus and associated product recommendation method |
CN106446943A (en) * | 2016-09-19 | 2017-02-22 | 大连海事大学 | Commodity correlation big data sparse network quick clustering method |
CN107220880A (en) * | 2017-05-26 | 2017-09-29 | 杭州再顾科技有限公司 | Method, equipment and server for showing commodity in shopping at network |
CN109242633A (en) * | 2018-09-20 | 2019-01-18 | 阿里巴巴集团控股有限公司 | A kind of commodity method for pushing and device based on bigraph (bipartite graph) network |
CN110009457A (en) * | 2019-04-09 | 2019-07-12 | 昆山古鳌电子机械有限公司 | A kind of commercial product recommending system based on big data |
Non-Patent Citations (1)
Title |
---|
卢丹;王君博;武森;: "电子商务中基于复杂网络社团发现的商品推荐研究", 工业技术创新, no. 01, pages 61 - 65 * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112052403A (en) * | 2020-09-11 | 2020-12-08 | 深圳市梦网视讯有限公司 | Commodity community classification-based link propagation method, system and equipment |
CN112052403B (en) * | 2020-09-11 | 2024-01-02 | 深圳市梦网视讯有限公司 | Commodity community classification-based link degree propagation method, system and equipment |
Also Published As
Publication number | Publication date |
---|---|
CN110992153B (en) | 2023-06-16 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
TWI740106B (en) | Product recommendation method and device | |
US10846775B1 (en) | Identifying item recommendations through recognized navigational patterns | |
CN108121737B (en) | Method, device and system for generating business object attribute identifier | |
US9959563B1 (en) | Recommendation generation for infrequently accessed items | |
CN108665333B (en) | Commodity recommendation method and device, electronic equipment and storage medium | |
US20220284457A1 (en) | Customer clustering using integer programming | |
TWI512653B (en) | Information providing method and apparatus, method and apparatus for determining the degree of comprehensive relevance | |
US9779441B1 (en) | Method for relevancy ranking of products in online shopping | |
CN110969516A (en) | Commodity recommendation method and device | |
CN107464132B (en) | Similar user mining method and device and electronic equipment | |
CN109711867B (en) | Shopping image construction marketing method and system based on audience big data | |
CN106708844A (en) | User group partitioning method and device | |
CN106055710A (en) | Video-based commodity recommendation method and device | |
CN102385727A (en) | ID-value assessment device, ID-value assessment system, and ID-value assessment method | |
CN108805598A (en) | Similarity information determines method, server and computer readable storage medium | |
CN112015998A (en) | Commodity recommendation method based on user portrait | |
CN104809637A (en) | Commodity recommending method and system realized by computer | |
CN108229826A (en) | A kind of net purchase risk class appraisal procedure based on improvement bayesian algorithm | |
CN113077317A (en) | Item recommendation method, device and equipment based on user data and storage medium | |
CN111626767B (en) | Resource data issuing method, device and equipment | |
CN112907334A (en) | Object recommendation method and device | |
CN111340566B (en) | Commodity classification method and device, electronic equipment and storage medium | |
CN113763089A (en) | Article recommendation method and device and computer-readable storage medium | |
CN110992153A (en) | Commodity recommendation method, system and equipment based on user attributes and commodity types | |
CN106204163B (en) | Method and device for determining user attribute characteristics |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
CB02 | Change of applicant information | ||
CB02 | Change of applicant information |
Address after: 518000 Guangdong city of Shenzhen province Nanshan District Guangdong streets high in the four Longtaili Technology Building Room 325 No. 30 Applicant after: Shenzhen mengwang video Co.,Ltd. Address before: 518000 Guangdong city of Shenzhen province Nanshan District Guangdong streets high in the four Longtaili Technology Building Room 325 No. 30 Applicant before: SHENZHEN MONTNETS ENCYCLOPEDIA INFORMATION TECHNOLOGY Co.,Ltd. |
|
GR01 | Patent grant | ||
GR01 | Patent grant |