Use of Machine Learning Techniques in Financial Forecasting

Abstract

Financial institutions play a crucial role in the acceleration of the economic stability and sustainability. In fact, one of their major sources of gain consists of debts provided to their customers in the form of loans. However, the risk that this service carries can lead to financial instability within these institutions. To avoid such incidents, an adequate credit approval process is needed through different approaches to forecast and assess delinquency and avoid falsification. The aim of this paper is to deal with credit risk management through the development of a model that classifies potential borrowers as good or bad credits. Furthermore, the paper reflects the importance of using machine learning techniques in the financial field. Historical observations of good and bad credits for a number of customers based on mixed attributes were used to train and test our data by carefully applying data preparation techniques and analysis (Univariate, Bivariate using ANOVA and Chi-Square tests), and KNN algorithm. Consequently, an accuracy test led us to an optimal K based on significant independent variables.