Applying Machine Learning Techniques to the Analysis and Prediction of Financial Data

Abstract

Data analysis and processing allow for acquiring competitive advantages both in the business and academic and research worlds. One of the sciences that carries out this analysis is machine learning, which has evolved with greater emphasis in recent years due to its advantages and applicability in different areas. Aware of the importance and current relevance of data management for industries, especially in the banking sector, this study applies supervised learning techniques to generate classification and prediction models by treating a set of data from an Ecuadorian financial institution. Different algorithms are compared, and each of the steps to follow in constructing the models is explained in detail. This allows the financial entity to classify its clients as VIPs or not with greater certainty, as well as to predict the investment amounts of the potential clients based on variables such as age, occupation, and among others. The main results show that the K-nearest neighbor algorithm with k = 5 is optimal for classification, while for prediction, the multilayer perceptron algorithm is the most favorable.