Digital marketing program design based on abnormal consumer behavior data classification and improved homomorphic encryption algorithm

Abstract

This article endeavors to delve into the conceptualization of a digital marketing framework grounded in consumer data and homomorphic encryption. The methodology entails employing GridSearch to harmonize and store the leaf nodes acquired post-training of the CatBoost model. These leaf node data subsequently serve as inputs for the radial basis function (RBF) layer, facilitating the mapping of leaf nodes into the hidden layer space. This sequential process culminates in the classification of user online consumption data within the output layer. Furthermore, an enhancement is introduced to the conventional homomorphic encryption algorithm, bolstering privacy preservation throughout the processing of consumption data. This augmentation broadens the applicability of homomorphic encryption to encompass rational numbers. The integration of the Chinese Remainder Theorem is instrumental in the decryption of consumption-related information. Empirical findings unveil the exceptional generalization performance of the amalgamated model, exemplifying an AUC (area under the curve) value of 0.66, a classification accuracy of 98.56% for online consumption data, and an F1-score of 98.41. The enhanced homomorphic encryption algorithm boasts attributes of stability, security, and efficiency, thus fortifying our proposed solution in facilitating companies’ access to precise, real-time market insights. Consequently, this aids in the optimization of digital marketing strategies and enables pinpoint positioning within the target market.