CN-121810332-B - Selection decision method based on dynamic consumption trend prediction and related products

CN121810332BCN 121810332 BCN121810332 BCN 121810332BCN-121810332-B

Abstract

The application discloses a selection decision method based on dynamic consumption trend prediction and related products. Firstly, global interactive users of a target brand party are obtained to obtain a plurality of target users, and a multi-mode behavior data set of each target user in a target time period is obtained to construct a user image library comprising a plurality of user semantic vectors. Obtaining user semantic vectors with time marks in a time sliding window to obtain a plurality of target semantic vectors, and performing time sequence dynamic semantic cluster analysis on the target semantic vectors to obtain a plurality of current semantic clusters. And identifying the number of users of the target users corresponding to each current semantic cluster, and calculating the trend score of each current semantic cluster by combining a trend scoring formula. And finally, screening target clusters based on trend scores of all the current semantic clusters, and generating a choice report by combining the similarity of the SKUs to be selected and the target clusters. The application can more sharply capture the real demands of consumers and dynamically predict the future consumption trend.

Inventors

CHEN BIYONG
FANG MIN
SHE ZHIYONG

Assignees

厦门南讯股份有限公司

Dates

Publication Date: 20260508
Application Date: 20260306

Claims (7)

1. A choice decision method based on dynamic consumption trend prediction, the method comprising: the method comprises the steps of obtaining a plurality of target users by global interactive users of a target brand side, and obtaining a multi-modal behavior data set of each target user in a target time period, wherein the multi-modal behavior data set comprises at least one behavior record, and each behavior record comprises stock unit SKU attributes, search keywords and customer service dialogue records generated based on different interactive behaviors of the users, and commodity evaluation or return reasons; constructing a user image library based on the multi-mode behavior data set corresponding to each target user, wherein the user image library comprises a plurality of user semantic vectors for describing consumption characteristics and product demand characteristics of each target user, and one user semantic vector corresponds to one behavior record; Obtaining a plurality of target semantic vectors by obtaining user semantic vectors with time marks in a time sliding window, and performing time sequence dynamic semantic cluster analysis on the plurality of target semantic vectors by adopting a density-based clustering algorithm to obtain a plurality of current semantic cluster; Identifying the number of users of the target users corresponding to each current semantic cluster, and calculating trend scores of the current semantic clusters based on a trend scoring formula and the number of users of the target users corresponding to the current semantic clusters, wherein the trend scores are used for evaluating development potential and outbreak degree of consumption trends corresponding to the current semantic clusters; screening target clusters based on the trend scores of the current semantic clusters, and generating a choice report by combining the similarity of the SKUs to be selected and the target clusters; The trend scoring formula is as follows: Strend=α×log(Vcurrent)+β×(Vcurrent-Vpast)/Vpast; Wherein Strend is trend score, α is scale weight, vcurrent is number of users of current semantic cluster, β is growth weight, vpast is number of users of last obtained history semantic cluster corresponding to current semantic cluster; The screening target clusters based on the trend scores of the current semantic clusters, and generating a choice report by combining the similarity of the SKUs to be selected and the target clusters, comprises the following steps: screening each current semantic cluster according to a screening rule and the trend score of each current semantic cluster to obtain at least one target cluster; obtaining the product semantic vectors of each SKU to be selected of the target brand party, and calculating the similarity between each product semantic vector and the clustering center of the target cluster; Generating the option report based on the similarity; one similarity corresponds to one SKU to be selected; the generating the option report based on the similarity includes: if the similarity is greater than or equal to a first similarity threshold, generating a production recommendation suggestion of the SKU to be selected corresponding to the similarity, wherein the production recommendation suggestion is used for suggesting direct production of the SKU to be selected and suggesting to increase the first stock quantity; generating an optimization adjustment suggestion if the similarity is smaller than a first similarity threshold and larger than or equal to a second similarity threshold, wherein the optimization adjustment suggestion is used for suggesting a research and development department to modify the design aiming at the missing pain points or only conduct small-batch money measurement; generating a desquamation suggestion if the similarity is smaller than a second similarity threshold value, wherein the desquamation suggestion is used for suggesting to directly desquamate or put into production for the SKU to be selected; integrating the production recommendation suggestion, the optimization adjustment suggestion and the elimination suggestion to generate the option report; Wherein the first similarity threshold is greater than the second similarity threshold.
2. The method of claim 1, wherein performing a time-sequential dynamic semantic cluster analysis on the plurality of target semantic vectors using a density-based clustering algorithm to obtain a plurality of current semantic clusters comprises: When the clustering execution stage is the first time sequence clustering, a density-based clustering algorithm is directly adopted to cluster the plurality of target semantic vectors to obtain a plurality of current semantic cluster; When the clustering execution stage is non-first time sequence clustering, taking the clustering center of each history semantic cluster obtained last time as an initial center point, and clustering the plurality of target semantic vectors based on each initial center point to obtain a plurality of current semantic clusters.
3. The method of claim 1, wherein the constructing a user image library based on the multimodal behavioral data sets corresponding to each target user comprises: Generating at least one first semantic reasoning prompt word based on at least one behavior record and a first preset instruction corresponding to each target user, and inputting the at least one first semantic reasoning prompt word into a large model for analysis to obtain at least one first semantic tag set, wherein the first semantic reasoning prompt word corresponds to one first semantic tag set which is used for describing the consumption characteristics and the product demand characteristics of the user, and each first semantic tag set at least comprises a user consumption intention, a user purchase scene, a user consumption style and a product pain point; Vectorizing a first semantic tag set corresponding to each target user to obtain at least one user semantic vector; And integrating, storing and marking time of all the user semantic vectors to obtain the user portrait library.
4. A method according to claim 3, characterized in that the method further comprises: when the target user corresponds to at least two first semantic tag sets, carrying out tag combination on the first semantic tag sets with at least N repeated tags to obtain combined tag sets, and deleting the original repeated first semantic tag sets, wherein N is a positive integer; vectorizing the first semantic tag set corresponding to each target user to obtain at least one user semantic vector, including: and vectorizing the first semantic tag sets and/or the combined tag sets corresponding to the target users to obtain at least one user semantic vector.
5. The method of claim 1, wherein the obtaining product semantic vectors for each candidate SKU of the target brand party comprises: the method comprises the steps of obtaining core attributes of each SKU to be selected, and converting the core attributes of each SKU to be selected into text descriptions to obtain structural product descriptions, wherein the core attributes are used for describing styles, formats, materials, functions and applicable scenes of the SKUs to be selected; Generating a second semantic reasoning prompt word based on the structured product description and a second preset instruction aiming at each structured product description, and inputting the second semantic reasoning prompt word into a large model for analysis to obtain a second semantic tag set, wherein the second semantic tag set is used for describing the core product characteristics of the SKU to be selected, and each second semantic tag set at least comprises a style, a model, a material, a function and an applicable scene; and vectorizing each second semantic tag set to obtain a plurality of product semantic vectors.
6. An item decision device based on dynamic consumption trend prediction, the device comprising: The system comprises a first acquisition unit, a second acquisition unit and a third acquisition unit, wherein the first acquisition unit is used for acquiring a plurality of target users from a global interactive user of a target brand party and acquiring a multi-modal behavior data set of each target user in a target time period, wherein the multi-modal behavior data set comprises at least one behavior record, and each behavior record comprises SKU attributes, search keywords and customer service dialogue records generated based on different interactive behaviors of the user and commodity evaluation or return reasons; The image library construction unit is used for constructing a user image library based on the multi-mode behavior data set corresponding to each target user, wherein the user image library comprises a plurality of user semantic vectors for describing consumption characteristics and product demand characteristics of each target user, one user semantic vector corresponds to one behavior record, and the user semantic vector is marked with a time mark which is the generation time of the behavior record corresponding to the user semantic vector; The second acquisition unit is used for acquiring the user semantic vectors with time marks in the time sliding window to obtain a plurality of target semantic vectors; the clustering unit is used for carrying out time sequence dynamic semantic cluster analysis on the plurality of target semantic vectors by adopting a density-based clustering algorithm to obtain a plurality of current semantic clusters; the user number identification unit is used for identifying the number of users of the target users corresponding to each current semantic cluster; the scoring unit is used for calculating the trend score of each current semantic cluster based on a trend scoring formula and the number of users of the target users corresponding to each current semantic cluster, wherein the trend score is used for evaluating the development potential and the outbreak degree of the consumption trend corresponding to the current semantic cluster; a cluster screening unit, configured to screen a target cluster based on the trend scores of the current semantic clusters; The first report generation unit is used for generating a choice report by combining the similarity of the SKUs to be selected and the target cluster; The trend scoring formula is as follows: Strend=α×log(Vcurrent)+β×(Vcurrent-Vpast)/Vpast; Wherein Strend is trend score, α is scale weight, vcurrent is number of users of current semantic cluster, β is growth weight, vpast is number of users of last obtained history semantic cluster corresponding to current semantic cluster; The cluster screening unit 507 is specifically configured to: screening each current semantic cluster according to a screening rule and the trend score of each current semantic cluster to obtain at least one target cluster; The first report generating unit 508 specifically includes: A third obtaining unit, configured to obtain a product semantic vector of each SKU to be selected of the target brand party; the similarity calculation unit is used for calculating the similarity between each product semantic vector and the clustering center of the target cluster; a second report generating unit configured to generate the option report based on the similarity; one similarity corresponds to one SKU to be selected; the second report generating unit is specifically configured to: if the similarity is greater than or equal to a first similarity threshold, generating a production recommendation suggestion of the SKU to be selected corresponding to the similarity, wherein the production recommendation suggestion is used for suggesting direct production of the SKU to be selected and suggesting to increase the first stock quantity; generating an optimization adjustment suggestion if the similarity is smaller than a first similarity threshold and larger than or equal to a second similarity threshold, wherein the optimization adjustment suggestion is used for suggesting a research and development department to modify the design aiming at the missing pain points or only conduct small-batch money measurement; generating a desquamation suggestion if the similarity is smaller than a second similarity threshold value, wherein the desquamation suggestion is used for suggesting to directly desquamate or put into production for the SKU to be selected; integrating the production recommendation suggestion, the optimization adjustment suggestion and the elimination suggestion to generate the option report; Wherein the first similarity threshold is greater than the second similarity threshold.
7. An item selection decision device based on dynamic consumption trend prediction, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor implements the item selection decision method based on dynamic consumption trend prediction as claimed in any one of claims 1-5 when executing the computer program.

Description

Selection decision method based on dynamic consumption trend prediction and related products Technical Field The application relates to the technical field of artificial intelligence, in particular to a selection decision method based on dynamic consumption trend prediction and a related product. Background Traditional customer relationship management (Customer Relationship Management, CRM) systems focus mainly on analysis of historical data, and can only remit what has been sold in the past, and basic business insights of hot-sell products, peak sales time and the like are extracted through collection and analysis of past sales data. However, such systems lack prospective research into market demand, nor do they deeply interpret the actual consumer demand. Traditional CRM can only clearly present past sales results, but does not have the capability of predicting future market trends and capturing consumer dynamic change demands. And the product demand iteration speed of modern consumers is high, the preference dimension is multiple, if enterprises can not accurately capture the demand signals changing in real time, new product development can deviate from market reality, and not only burst products with agreeable trends are difficult to build, but also the lost sales can be caused by the positioning of the new products and the dislocation of market demands, so that the ineffective investment of research and development and production resources is caused. Therefore, how to capture the real demands of consumers and dynamically predict the consumption trend is a technical problem that the skilled person is urgent to solve. Disclosure of Invention Based on the problems, the application provides a selection decision method based on dynamic consumption trend prediction and related products, which can capture the real demands of consumers and dynamically predict the consumption trend. The embodiment of the application discloses the following technical scheme: a choice decision method based on dynamic consumption trend prediction, the method comprising: the method comprises the steps of obtaining a plurality of target users by global interactive users of a target brand side, and obtaining a multi-modal behavior data set of each target user in a target time period, wherein the multi-modal behavior data set comprises at least one behavior record, and each behavior record comprises stock unit SKU attributes, search keywords and customer service dialogue records generated based on different interactive behaviors of the users, and commodity evaluation or return reasons; constructing a user image library based on the multi-mode behavior data set corresponding to each target user, wherein the user image library comprises a plurality of user semantic vectors for describing consumption characteristics and product demand characteristics of each target user, and one user semantic vector corresponds to one behavior record; Obtaining a plurality of target semantic vectors by obtaining user semantic vectors with time marks in a time sliding window, and performing time sequence dynamic semantic cluster analysis on the plurality of target semantic vectors by adopting a density-based clustering algorithm to obtain a plurality of current semantic cluster; Identifying the number of users of the target users corresponding to each current semantic cluster, and calculating trend scores of the current semantic clusters based on a trend scoring formula and the number of users of the target users corresponding to the current semantic clusters, wherein the trend scores are used for evaluating development potential and outbreak degree of consumption trends corresponding to the current semantic clusters; and screening target clusters based on the trend scores of the current semantic clusters, and generating a choice report by combining the similarity of the SKUs to be selected and the target clusters. In one possible implementation manner, the performing time-sequential dynamic semantic cluster analysis on the plurality of target semantic vectors by using a density-based clustering algorithm to obtain a plurality of current semantic clusters includes: When the clustering execution stage is the first time sequence clustering, a density-based clustering algorithm is directly adopted to cluster the plurality of target semantic vectors to obtain a plurality of current semantic cluster; When the clustering execution stage is non-first time sequence clustering, taking the clustering center of each history semantic cluster obtained last time as an initial center point, and clustering the plurality of target semantic vectors based on each initial center point to obtain a plurality of current semantic clusters. In one possible implementation manner, the constructing a user image library based on the multimodal behavior data set corresponding to each target user includes: Generating at least one first semantic reasoning prompt word based on at least o