Computational analysis of user big data of electric vehicles based on SVM and dynamic planning

Abstract

In order to mine new energy vehicle users according to the user’s scoring table, first use the Spearman correlation coefficient method to calculate the correlation coefficients of the three brands to determine the significance of each brand; Secondly, establish an SVM model and use the trained SVM model to calculate the purchase probability of the population to be predicted. The accuracy rate of the result of brand one is 96%, the accuracy rate of brand two is 92.2%, and the accuracy rate of brand three is 88%; Finally, the dynamic programming method is used to calculate how the service intensity is allocated to the eight product indicators, so as to achieve the optimal customer plan. The final result showed that: for the joint venture brand, the battery technical performance, comfort, economy, and power were increased by 1%, and the willingness to be obtained by using the SVM model was changed from not buying to buying; Regarding independent brands, battery technical performance, comfort, safety performance, economy, power performance and driving control performance, appearance and interior decoration, each increase 0.83% of user satisfaction, and the prediction result becomes a purchase; For new power brands, battery technical performance, comfort, economy, power performance and driving control performance each increase user satisfaction by 1%, the prediction result will become a purchase.