Subgroup Discovery Methods Development for Individual Treatment Effect Estimation

The task of treatment effect estimation is becoming more and more popular and promising in various areas of human life. For example, the ability to get an accurate assessment of the power of influence of a new medicine for a certain group of people will allow medicine to more effectively fight diseases. And for marketers, solving this problem will create more effective campaigns to retain old customers and maintain their loyalty. However, there is still no universal and optimal method for solving this problem. Therefore, in this paper, an attempt will be made to develop and test a new approach based on Subgroup Discovery methods to assess the individual treatment effect. Testing the effectiveness of the approach will take place on the real data of one large retail network. In the first chapter, the statement of the problem and the analysis of the subject area of the problem of treatment effect estimation are provided, as well as the rationale for the approach used. The second chapter analyzes the main elements that make up Subgroup Discovery methods and conducts an experiment on synthetic data using the approaches considered. The third chapter is devoted to the application of the proposed approach on real data, assessing the quality of the results and summarizing the whole work. Final qualification work includes an abstract, introduction, three main chapters, conclusion, list of sources used. Volume: 52 pages.