Using Gradient Boosting Machines to Extrapolate the Somaliland Consumer Price Index: An Analysis of Food and Non-Food Indices

Abstract

With the use of two prediction models such as ARIMAX and Gradient Boosting Machines, this study seeks to forecast the Consumer Price Index (CPI) in Somaliland (GBMs). The driving force is the CPI's crucial role in determining economic policy as well aNos how it affects financial planning and decision-making. The goal is to evaluate how well GBMs and conventional ARIMAX perform in predicting the CPI while taking into account predictors like FOOD and NON-FOOD. Using historical CPI, FOOD, and NON-FOOD data, the study's methodology combines gradient boosting regression for GBMs and ARIMA models with exogenous variables for ARIMAX. With lower MSE (ARIMAX: 0.69967 vs. GBMs: 0.0485), RMSE (ARIMAX: 0.8364 vs. GBMs: 0.2202), and greater R-squared (ARIMAX: 28.4% vs. GBMs: 91.6%), the results demonstrate that GBMs perform better than ARIMAX in terms of predicting accuracy. Furthermore, the GBMs model's predicted CPI values for Somaliland over the following 24 months indicate a steady increasing trend, supporting the model's applicability to economic planning. According to the study's findings, GBMs are a better fit for Somaliland's CPI forecasting, which will improve economic planning and policymaking. For more reliable forecasts, it is advised to include real-time economic variables and modify the model continuously.