Development of Methology for Applying Machine Learning Methods to Solve Marketing Problems in the Telecommunications Business

This work is dedicated to the development of a methodology for applying machine learning methods to solve marketing problems in the telecommunication business. The relevance of this subject and practical importance of the existence of such methodology are given in this research. During the implementation of the main stages of the work, key points of the work are solved, and that is why it becomes possible to achieve the main purpose of the research: to develop a universal methodology that allows to use machine learning methods for solving various marketing tasks within telecommunications companies. There are three stages of the research. During the first stage of the study, a review of available algorithms for machine learning was made, as well as tools that allow the use of optimal methods of machine learning. The selection of the optimal software for constructing the model using the expert decision support system was also carried out. During the second stage of the work, there was a description of data marts created to store information about various aspects related to the subscriber's life within the telecommunication company's services. Also, the principle of selecting the most significant predictors for their subsequent use in the development of the predictive model was demonstrated. During the third stage of the work, a predictive model was designed to solve one of the telecommunication operator's marketing tasks - forecasting subscribers who are inclined to purchase additional services. In the framework of this process, a description of the principle of data processing was given, the formation of target metrics to assess the quality of the model, as well as the approach to selecting optimal parameters and interpreting the results. All stages of design and development were based on current information from the subject literature and other specialized sources, and were also supported by screenshots, tables and diagrams.