The application of machine learning in the classification and classification of securities and futures customers

How to identify high-value customers among the massive customer base and achieve precise marketing and service is the current challenge facing securities and futures companies. The traditional method of dividing customer groups according to the amount of assets is more based on experience and not accurate enough. The goal of the research is to explore if machine learning algorithms can solve the above problem. In this study, a K-means clustering model is built to categorize individual customers into different groups based on their behavior. The Elbow method and Gap Statistics are used to determine 7 as the best number of clusters, and the corresponding K-means model is able to group customers in a more accurate way with regard to client total contribution to the firm’s revenue. Later, a gradient boost algorithm on a decision tree is developed to quantitatively score customers based on a weighted average of various dimensions. The 2 most important dimensions are net retained transaction fees and assets according to the model. These 2 models can help improve the accuracy of locating key customers compared to traditional methods.