Determining the Popularity of Geolocation for ATM Placement

This work consists of 3 chapters, 32 pages, 23 figures, 4 tables, 25 sources used. The goal of this work is to create a service for assessing the potential location of an ATM at given coordinates using automation of data enrichment, training a machine learning model (while maintaining the possibility of manual and variable launch of model training), as well as a clear and sufficient user interface. The first chapter examines the original data set that was originally collected and used for the corresponding competition. An initial exploratory analysis of the data and its further enrichment were carried out. The mechanism for collecting data, its types and organization are described. The second describes the main modeling approaches using machine learning models and their descriptions, as well as building basic models and examining the obtained data. Based on all of the above, a comparison is made and the transition to the use of more complex models is carried out. Also their subsequent comparison, assessment of accuracy and speed, fine tuning of model parameters. Then the final model is selected, which will be used directly in the service. The third chapter describes the complex operation of the entire service, as well as its individual components and tools used in the service. It also describes the system requirements for the service, versions of the various libraries used, and the results of speed tests of the service and its individual components.