Method for solving sales promotion commodity combination based on commodity influence analysis
Method for solving sales promotion commodity combination based on commodity influence analysis
 CN 106,228,411 B
 Filed: 07/31/2016
 Issued: 08/31/2021
 Est. Priority Date: 07/31/2016
 Status: Active Grant
First Claim
1. A method for solving a promotion commodity combination based on commodity influence analysis is characterized by comprising the following steps:
 (1) constructing a commodity network;
the method for constructing the commodity network comprises the following steps;
(1a) inputting the combined information of the commodities;
(1b) calculating the relevance between the commodities by using an apriori algorithm;
(1c) representing the commodities with the support degree not less than the minimum support degree by using the nodes;
if the strength of the relevance between the two commodities is not less than the minimum confidence coefficient, the two commodity nodes are considered to have a mutual influence relationship, then the influence relationship is represented by a directed edge in the network, and the weight value of the edge represents the influence strength between the commodities;
(2) dividing communities;
the community division is to divide the commodity network into a plurality of communities by using a directed graph rapid hierarchical clustering algorithm according to the characteristics of modules existing among commodities;
(3) calculating the influence gain of each commodity;
the method for calculating the influence gain of each commodity is to use a social network independent cascade model to model commodity influence propagation and calculate the influence gain of the commodity;
(4) calculating the community score of each community for the communities obtained in the step (2);
calculating a community score (score) of each community for the communities obtained in the step (2) as shown in formula (1);
num _ node is the number of nodes of the community, seed _ num is the number of currently selected seeds, the community score is based on the concept, the number of the affected nodes of each seed node of the community is the same, the seed nodes in the community can affect all the nodes, if the number of the selected seeds of the community is more, the number of the affected nodes of each node on average is reduced, namely the score of the community is reduced;
if the influence gains of the two nodes are the same, selecting a seed node with a large community score, wherein the potential influence of the seed node in practice may be larger than that of the seed node with a small community score;
Chinese PRB Reexamination
Abstract
The invention discloses a method for solving a sales promotion commodity combination based on commodity influence analysis, and belongs to the field of shopping basket analysis. The method comprises the following implementation steps: (1) and constructing a directed commodity network with the right according to the commodity combination information. (2) And carrying out community division on the commodity network. (3) The impact gain value for each article of merchandise is calculated. (4) A community score is calculated for each community. (5) And selecting a sales promotion commodity according to the community score and the influence gain value. (6) For the community which selects the sales promotion commodity, the community score of the community and the influence gain of the commodity are recalculated. And (5) if the number of the selected sales promotion commodities is less than a preset value k, continuing to select the sales promotion commodities in the step (5), otherwise, outputting the k sales promotion commodities. The method takes the selected commodities with the largest influence as the sales promotion commodities, thereby effectively solving the selection problem of the sales promotion commodity combination. In the application scene of the auxiliary promotion decision, the method has wide application prospect.
2 Claims

1. A method for solving a promotion commodity combination based on commodity influence analysis is characterized by comprising the following steps:

(1) constructing a commodity network; the method for constructing the commodity network comprises the following steps; (1a) inputting the combined information of the commodities; (1b) calculating the relevance between the commodities by using an apriori algorithm; (1c) representing the commodities with the support degree not less than the minimum support degree by using the nodes;
if the strength of the relevance between the two commodities is not less than the minimum confidence coefficient, the two commodity nodes are considered to have a mutual influence relationship, then the influence relationship is represented by a directed edge in the network, and the weight value of the edge represents the influence strength between the commodities;(2) dividing communities; the community division is to divide the commodity network into a plurality of communities by using a directed graph rapid hierarchical clustering algorithm according to the characteristics of modules existing among commodities; (3) calculating the influence gain of each commodity; the method for calculating the influence gain of each commodity is to use a social network independent cascade model to model commodity influence propagation and calculate the influence gain of the commodity; (4) calculating the community score of each community for the communities obtained in the step (2); calculating a community score (score) of each community for the communities obtained in the step (2) as shown in formula (1); num _ node is the number of nodes of the community, seed _ num is the number of currently selected seeds, the community score is based on the concept, the number of the affected nodes of each seed node of the community is the same, the seed nodes in the community can affect all the nodes, if the number of the selected seeds of the community is more, the number of the affected nodes of each node on average is reduced, namely the score of the community is reduced;
if the influence gains of the two nodes are the same, selecting a seed node with a large community score, wherein the potential influence of the seed node in practice may be larger than that of the seed node with a small community score;


2. A method for solving a promoted merchandise combination based on merchandise influence analysis according to claim 1, wherein the method for selecting a promoted merchandise from the candidate promoted merchandise of all communities in the step (5b) is as follows:
suppose the candidate promotional item for community a is a1, the impact gain value is a _ influ, and the community score currently for community a is a _ score;
similarly, suppose that the candidate promotion item of community B is B1, the impact gain value is B _ influ, and the current community score of community B is B _ score;
suppose that a promotional item is to be selected from the candidate promotional items of communities a and B;
if a _ influ is greater than b _ influ, then a1 is selected as the promotional item;
if a _ influ is equal to B _ influ, then the community score needs to be considered, if B _ score is greater than A _ score, then B1 is selected as the promotional item.
Specification(s)