This article focuses on user evaluation procedures for Chinese investment companies. It uses a clustering process to identify different categories of users and uses regression to evaluate and provide a score. These two factors serve as the basis for this organization to develop unique tactics for its various clients, which benefit both the company and its users. This project aims to serve existing customers better and reduce the risk of losing critical new customers. Clean the data to remove inactive accounts and outliers, and use logarithmic processing to limit the impact of high monetary values. This is due to the significant relationship between variables. Running algorithms on raw data is difficult. So, to condense many data dimensions, we use component analysis to define the three dimensions (amount, transaction amount, profit) that reflect consumer information. For clustering, we use the commonly known K-means clustering algorithm. Customers are classified into four categories using an angled approach. The four groups formed include the high trading frequency group, the large and profit group, the large and loss group, and the majority trading contract group. Use a regression Tree to perform regression based on reduced dimensions and their contributions. This model achieves 97% accuracy, indicating that the financial characteristics of the users are fundamental to the business. Additional discussion validates the clustering results using classification and regression methods on several contributing variables to provide further insight into each dimension.
Keywords: K-means clustering, Quantitative Analysis, feature importance, regression tree
