Application of the random forest method for purchasing power analysis

The article considers the application of the random forest machine learning method in a real business situation. The aim of the work set the task: having a database of clients collected when registering them in a mega market, quantitatively estimated store marketers for each customer manually, to develop a system based on machine learning, which is able to automatically classify customers by their solvency. Achieving this aim assumed the solution of the following tasks: 1) on the basis of the collected customer database and their creditworthiness groups, to analyze and determine which criteria specified by clients at the time of registration most affect the perceived solvency; 2) normalize collected customer data to fit the selected machine learning method; 3) train the model and test its work on clients, already with the specified solvency groups; 4) improve the system’s accuracy. The article reveals the advantages of this algorithm and describes the entire development process: from zero to a correctly functioning model that can automatically classify new store customers according to their solvency. The final accuracy of the model was approx. 81 %. Implementation of the proposed model enables to automate the business process of buyers’ solvency analysis and increase the time and human resources efficiency.