LEVERAGING MACHINE LEARNING FOR FINANCIAL INCLUSION PREDICTION IN NIGERIA: A DEVELOPMENT FINANCE APPROACH

Abstract

Financial inclusion remains an essential enterprise in many developing economies, especially in sub-Saharan Africa where a large proportion of the population lacks access to formal economic services. This study presents a machine learning framework for predicting economic inclusion using demographic, social and monetary records. Three models K-Nearest Neighbors (KNN), Random Forest (RF), and an Ensemble vote casting classifier were developed and evaluated using Zindi Nigerian financial inclusion dataset, the dataset comprises 33,612 samples, which were partitioned into training and testing subsets in a 70:30 ratio. A total of 23,525 samples were allocated for training, while 10,087 samples were used for testing, coping with missing values in data preprocessing, through normalization, feature selection and data cleaning. Models were evaluated using Precision, Accuracy, AUC-ROC and F1-score metrics. The experimental result shows that the ensemble model executed a superior overall performance with 92% accuracy, outperforming the individual model.

Keywords: Prediction, Model, Credit Risk.