Search

CN-122022955-A - Digital marketing recommendation method and system based on big data

CN122022955ACN 122022955 ACN122022955 ACN 122022955ACN-122022955-A

Abstract

The invention discloses a digital marketing recommendation method and system based on big data, and relates to the technical field of commodity recommendation; the method comprises the steps of obtaining original data of an e-commerce platform, preprocessing to obtain a structured data set, carrying out fuzzy clustering on a scoring matrix, generating and iteratively updating a membership matrix, reducing the influence of initial clustering deviation, calculating clustering attribution by combining with user geographic information, correcting scores to obtain a final scoring matrix according to the final scoring matrix, inputting the final scoring matrix, a user and commodity feature vector into a recommendation model, outputting commodity recommendation results, clustering the user with multiple membership by adopting fuzzy clustering, adapting to complex overlapping features of user preference under sparse data of the e-commerce, reducing the initial deviation by iterative optimization membership, combining with geographic information, and improving clustering fitting degree and distinguishing degree, wherein the commodity recommendation model effectively solves the problem of lower commodity recommendation result accuracy under the condition of reducing calculation complexity.

Inventors

  • Meng Sigang
  • ZENG YONG

Assignees

  • 天下广宣(杭州)网络科技有限公司

Dates

Publication Date
20260512
Application Date
20260205

Claims (10)

  1. 1. A digital marketing recommendation method based on big data, the method comprising: The method comprises the steps of obtaining original data of an e-commerce platform, preprocessing the original data to obtain a structured data set, wherein the structured data set comprises a scoring matrix, a user characteristic vector, a commodity characteristic vector and a commodity public scoring list; performing fuzzy clustering on the scoring matrix to generate an initial membership matrix, and performing iterative updating on the initial membership matrix to obtain a final membership matrix; collecting geographic information of a user, and calculating to obtain a clustering attribution result of the user according to the geographic information of the user and a final membership matrix; Correcting the original scores of the users and the public scores of the commodities according to the clustering attribution results of the users to obtain corrected scores of the users and the commodities, and presetting the corrected scores to obtain a final scoring matrix; And inputting the final scoring matrix, the user feature vector and the commodity feature vector into a commodity recommendation model to output commodity recommendation results, and uploading the commodity recommendation results to a cloud.
  2. 2. The digitized marketing recommendation method based on big data of claim 1, wherein iteratively updating the initial membership matrix to obtain a final membership matrix comprises: Dividing the scoring matrix according to rows to obtain scoring vectors of each user, and generating an initial membership matrix according to the scoring vectors; calculating an iterated clustering center through preset parameters according to the scoring vectors of all the users and the initial membership corresponding to each cluster; Calculating the distance from each user scoring vector to each clustering center according to the iterated clustering centers, and calculating the iterated membership according to the distance relation and preset parameters; And calculating a difference value between the membership degree after iteration and the membership degree after the previous iteration, judging according to the difference value and an iteration termination threshold value, executing iteration continuation or termination according to a judgment result, and outputting a final membership degree matrix.
  3. 3. The digital marketing recommendation method based on big data according to claim 1, wherein the clustering attribution result of the user is obtained by calculating according to the geographic information and the final membership matrix of the user, comprising: Step 1, selecting a plurality of users in a target area in geographic information, and extracting membership of the users in the area to a target cluster center from a final membership matrix, wherein the target cluster center is any one of a plurality of cluster centers, and the target area is any area position in the geographic information; Step 2, the membership degrees are arranged in an ascending order to obtain a membership degree sequence, and the membership degrees with preset proportions are selected from the membership degree sequence to serve as membership degree thresholds; Step 3, comparing the membership of the user in the target area in the step 1 with a membership threshold, if the membership of the user is greater than the membership threshold, the user belongs to a target cluster center, and if the membership of the user is not greater than the membership threshold, the user does not belong to the target cluster center; And 4, sequentially executing the steps 1 to 3 aiming at each clustering center to obtain clustering attribution results of all users.
  4. 4. The method of claim 1, wherein the step of obtaining a final scoring matrix from the correction scoring preset operation comprises: Calculating original scores and commodity public scores belonging to the target clustering center users through preset correction parameters to obtain correction scores; Calculating a preference judgment index according to the original score of the user and the commodity public score, and setting a preference grade according to the preference judgment index; And counting the number of users of each preference level, calculating the preference probability of each preference level, and constructing a final scoring matrix according to the correction scores, the preference levels and the preference probabilities.
  5. 5. The digital marketing recommendation method based on big data of claim 1, wherein the commodity recommendation model works on the principle that: Acquiring input data, inputting the input data to a full connection layer to output a1 st target feature, inputting the 1 st target feature to a CNN module to obtain a 2 nd target feature, activating a ReLU function to obtain a 3 rd target feature, sequentially inputting the 2 nd target feature to two GNN modules to obtain a4 th target feature, and splicing and fusing the 1 st target feature, the 3 rd target feature and the 4 th target feature to obtain a 5 th target feature; Processing the 5 th target feature through an LDA module to obtain a 6 th target feature, respectively inputting the 6 th target feature into a channel attention module to obtain a1 st channel feature and a2 nd channel feature, respectively fusing the 1 st channel feature and the 2 nd channel feature with the 6 th target feature, and respectively inputting the 1 st channel feature and the 2 nd channel feature into a space attention module to obtain a3 rd channel feature and a4 th channel feature; And fusing the 1 st channel characteristic with the 3 rd channel characteristic to obtain a 5 th channel characteristic, fusing the 2 nd channel characteristic with the 4 th channel characteristic to obtain a 6 th channel characteristic, splicing the 5 th channel characteristic and the 6 th channel characteristic, sequentially inputting the spliced 5 th channel characteristic and the 6 th channel characteristic into an LSA module and a full-connection layer, and activating the LSA module through a Sigmoid function to output a recommendation result.
  6. 6. A big data based digital marketing recommendation system, the system comprising: the system comprises a preprocessing module, a processing module and a processing module, wherein the preprocessing module is used for acquiring original data of an e-commerce platform and preprocessing the original data to obtain a structured data set, wherein the structured data set comprises a scoring matrix, a user characteristic vector, a commodity characteristic vector and a commodity public scoring list; The clustering module is used for performing fuzzy clustering on the scoring matrix to generate an initial membership matrix, and performing iterative updating on the initial membership matrix to obtain a final membership matrix; the result calculation module is used for collecting the geographic information of the user and calculating to obtain a clustering attribution result of the user according to the geographic information of the user and the final membership matrix; The correction module is used for correcting the original scores of the users and the public scores of the commodities according to the clustering attribution results of the users to obtain corrected scores of the users and the commodities, and presetting the corrected scores to obtain a final scoring matrix; and the uploading module is used for inputting the final scoring matrix, the user feature vector and the commodity feature vector into the commodity recommendation model to output commodity recommendation results and uploading the commodity recommendation results to the cloud.
  7. 7. The big data based digital marketing recommendation system of claim 6, wherein the clustering module comprises: the splitting module is used for splitting the scoring matrix according to rows to obtain scoring vectors of each user, and generating an initial membership matrix according to the scoring vectors; The first calculation module is used for calculating an iterated cluster center through preset parameters according to the scoring vectors of all the users and the initial membership corresponding to each cluster; The second calculation module is used for calculating the distance from each user scoring vector to each clustering center according to the iterated clustering centers and calculating the iterated membership according to the distance relation and preset parameters; The judging module is used for judging according to the difference value and the iteration termination threshold value by calculating the difference value between the membership degree after iteration and the membership degree after the previous iteration, executing iteration continuation or termination according to the judging result, and outputting a final membership degree matrix.
  8. 8. The big data based digital marketing recommendation system of claim 6, wherein the results calculation module comprises a home diagnostic module, wherein: The attribution diagnosis module is used for executing the steps 1 to 4, and the process is as follows: Step 1, selecting a plurality of users in a target area in geographic information, and extracting membership of the users in the area to a target cluster center from a final membership matrix, wherein the target cluster center is any one of a plurality of cluster centers, and the target area is any area position in the geographic information; Step 2, the membership degrees are arranged in an ascending order to obtain a membership degree sequence, and the membership degrees with preset proportions are selected from the membership degree sequence to serve as membership degree thresholds; Step 3, comparing the membership of the user in the target area in the step 1 with a membership threshold, if the membership of the user is greater than the membership threshold, the user belongs to a target cluster center, and if the membership of the user is not greater than the membership threshold, the user does not belong to the target cluster center; And 4, sequentially executing the steps 1 to 3 aiming at each clustering center to obtain clustering attribution results of all users.
  9. 9. The big data based digital marketing recommendation system of claim 6, wherein the correction module comprises: The third calculation module is used for calculating the original scores and commodity public scores of the users belonging to the target clustering center through preset correction parameters to obtain correction scores; The grade setting module is used for calculating a preference judgment index according to the original scores of the users and the commodity public scores and setting preference grades according to the preference judgment index; The construction module is used for counting the number of users of each preference level, calculating the preference probability of each preference level and constructing a final scoring matrix according to the correction scores, the preference levels and the preference probabilities.
  10. 10. The digital marketing recommendation system based on big data of claim 6, wherein the uploading module comprises a commodity recommendation model working principle: Acquiring input data, inputting the input data to a full connection layer to output a1 st target feature, inputting the 1 st target feature to a CNN module to obtain a 2 nd target feature, activating a ReLU function to obtain a 3 rd target feature, sequentially inputting the 2 nd target feature to two GNN modules to obtain a4 th target feature, and splicing and fusing the 1 st target feature, the 3 rd target feature and the 4 th target feature to obtain a 5 th target feature; Processing the 5 th target feature through an LDA module to obtain a 6 th target feature, respectively inputting the 6 th target feature into a channel attention module to obtain a1 st channel feature and a2 nd channel feature, respectively fusing the 1 st channel feature and the 2 nd channel feature with the 6 th target feature, and respectively inputting the 1 st channel feature and the 2 nd channel feature into a space attention module to obtain a3 rd channel feature and a4 th channel feature; And fusing the 1 st channel characteristic with the 3 rd channel characteristic to obtain a 5 th channel characteristic, fusing the 2 nd channel characteristic with the 4 th channel characteristic to obtain a 6 th channel characteristic, splicing the 5 th channel characteristic and the 6 th channel characteristic, sequentially inputting the spliced 5 th channel characteristic and the 6 th channel characteristic into an LSA module and a full-connection layer, and activating the LSA module through a Sigmoid function to output a recommendation result.

Description

Digital marketing recommendation method and system based on big data Technical Field The invention belongs to the technical field of commodity recommendation, and particularly relates to a digital marketing recommendation method and system based on big data. Background The current e-commerce industry user scale and commodity category are in explosive growth, the traditional recommendation mode depends on single-dimension data and is easy to be restricted by problems of data sparseness, scoring deviation and the like, and the accuracy and the conversion rate are low. Massive user behaviors, commodity attributes, geographic information and other big data resources are not fully mined, and potential user preferences and group characteristics are difficult to capture. Under the background, the digital commodity recommendation technology based on big data is required to be accurately matched with supply and demand, so that user experience and platform operation efficiency are improved. The prior publication number is CN118365431A, which discloses a commodity recommendation method and a commodity recommendation system based on big data of an e-commerce platform, wherein the commodity recommendation method and the commodity recommendation system comprise the steps of analyzing visual information of commodity pictures based on seller commodity pictures, analyzing commodity characteristics and consumer emotion by combining commodity description and user comments, and generating commodity characteristic identification information. According to the invention, the commodity label accuracy and integrity are enhanced by extracting commodity characteristics and consumer emotion information and combining visual and text information, the descriptive and pertinence of a label system of the commodity are optimized, and the commodity demand is predicted by integrating interaction data of a user and a commodity label database and combining seasonal change and historical sales data. The method predicts the commodity demand by combining commodity description and user comments and extracting commodity characteristics and consumer emotion information, but the situation of sparse data can occur for some established electronic commerce platforms, and under the situation, the characteristic data of merchants and users are difficult to support accurate prediction of commodity marketing, and the problem that the accuracy of the recommendation result output by the traditional commodity recommendation model is lower is solved. Disclosure of Invention The invention aims to solve the problems that accurate prediction of commodity marketing is difficult to support in a data sparse scene and the accuracy of a recommendation result output by a traditional commodity recommendation model is low, and provides a digital marketing recommendation method and system based on big data. In a first aspect of the present invention, a digital marketing recommendation method based on big data is first provided, the method includes: The method comprises the steps of obtaining original data of an e-commerce platform, preprocessing the original data to obtain a structured data set, wherein the structured data set comprises a scoring matrix, a user characteristic vector, a commodity characteristic vector and a commodity public scoring list; performing fuzzy clustering on the scoring matrix to generate an initial membership matrix, and performing iterative updating on the initial membership matrix to obtain a final membership matrix; collecting geographic information of a user, and calculating to obtain a clustering attribution result of the user according to the geographic information of the user and a final membership matrix; Correcting the original scores of the users and the public scores of the commodities according to the clustering attribution results of the users to obtain corrected scores of the users and the commodities, and presetting the corrected scores to obtain a final scoring matrix; And inputting the final scoring matrix, the user feature vector and the commodity feature vector into a commodity recommendation model to output commodity recommendation results, and uploading the commodity recommendation results to a cloud. The method comprises the steps of generating a structured data set by preprocessing original data, relieving the problems of scattered data dimension and insufficient effective association, carrying out fuzzy clustering iteration to optimize a membership matrix, determining user clustering attribution by combining geographic information, compensating the defect of lack of user preference characteristics under sparse data, correcting scores according to clustering results, weakening abnormal interference, improving reliability of the scoring matrix, inputting multidimensional data into a recommendation model and uploading the multidimensional data to the cloud, and effectively improving accuracy and stability of commodity recommen