FORECASTING REAL PER CAPITA NATIONAL DISPOSABLE INCOME OF NIGERIA USING TIME SERIES AND MACHINE LEARNING MODELS

Abstract

This study aims to forecast the real per capita national disposable income of Nigeria using a combination of time series and machine learning models. We employ the seasonality adjusted autoregressive integrated moving average (SARIMA) time series model, a classical time series model, and Prophet, a modern machine learning approach, to analyze historical data on the real per capita national disposable income of Nigeria for the period 2010-2018. The models are compared in terms of accuracy, robustness, and ease of interpretation. Our findings clearly revealed that while the Prophet model is easy to use and fine-tune, the seasonality adjusted autoregressive integrated moving average (SARIMA) time series model was more effective and efficient in modeling the data as well as provided more accurate forecasted values. Our findings provide insights into the effectiveness of these methods for forecasting economic indicators in Nigeria. This study contributes to the literature on economic forecasting and provides practical guidance for policymakers and analysts.