Application of Improved k-means Algorithm in E-commerce Data Processing

Accurate recommendation processing for a large number of e-commerce products can play a role in increasing e-commerce sales and improving the user's consumption experience. This study uses genetic algorithm, coefficient of variation method to design an improved k-means algorithm, and design an improved singular value decomposition ++ algorithm, so as to construct an e-commerce product data recommendation model. The model uses the improved singular value decomposition ++ algorithm to extract the hidden features of the data, and the improved k-means algorithm to realize the recommendation of the products. The performance test results revealed that when the number of recommendations was 15, the area under the recommended precision, recall, and receiver operating characteristic curves of the designed recommendation model were 85%, 87%, and 0.83, respectively, which were higher than all the ablation experiment comparison models and advanced recommendation models. The average computational time consumption of ISVD++_I_k-means, RLRA, TRRA, and CF models were 54.2s, 73.8s, 83.3s, and 58.7s, respectively. Among them, ISVD++_I_k-means consumed less time, but the computational memory consumption of the designed model was in the worse level among all the comparison models. The test results demonstrate that, although there are certain drawbacks in terms of resource consumption, the recommendation model developed in this study can successfully increase the efficiency and quality of recommendations. The research results are beneficial to provide reference for e-commerce platforms to design more efficient recommendation models.