STATISTICAL MODELING AND FORECASTING BANK DEPOSIT DATA USING RANDOM FORESTS

In the field of financial and foreign exchange markets, large amounts of data of various nature are accumulated. Extracting essential information from this data aiming to provide a base for taking proper management decisions of the banks executive boards, companies and other financial institutions is an important practical task. The process of solving problems related to the processing of financial data in the era of ubiquitous digitalization is increasingly achieved with the application of the most modern and powerful mathematical tools and the techniques and algorithms developed on their basis. The purpose of this study is to analyze and forecast real data from the currency deposits of Bulgarian citizens. The data are in the form of a univariate time series on a monthly basis. They are modeled using the powerful Random forests (RF) machine learning method. Predictive RF models were created and tested, achieving up to 98% data matching. The models are applied to short-term forecasts, simulating the prediction of future deposit values.