Publication Details
Abstract
This study examines the development of the food industry in the Kashkadarya region of Uzbekistan by integrating classical regression analysis with the Random Forest (RF) algorithm. The research aims to identify the key economic factors influencing food industry development, evaluate their relationships with production performance, and compare the explanatory and predictive capabilities of conventional econometric and machine learning approaches. Multiple regression analysis is applied to estimate the direction, magnitude, and statistical significance of factor effects, whereas Random Forest is employed to capture potential nonlinear relationships and assess the relative importance of predictors. The performance of the models is evaluated using R², MAE, and RMSE metrics. Using annual data for 2011–2024, the estimated regression model explains approximately 96.8% of the variation in food production (R² = 0.9677; adjusted R² = 0.9580; F = 99.86). Investment in food production (coefficient = 22.5863, p = 0.0016) and the number of food manufacturing enterprises (coefficient = 693.7199, p = 0.0313) are statistically significant, whereas employment is not (p = 0.4507). In the Random Forest model, employment, investment, and the number of enterprises account for approximately 40.4%, 33.8%, and 25.8% of feature importance, respectively, suggesting nonlinear effects or interactions not captured by the linear specification. The findings provide a methodological basis for identifying the major drivers of regional food industry development, improving forecasting accuracy, and supporting evidence-based decisions aimed at strengthening investment activity, production capacity, and the sustainable development of the food industry in the Kashkadarya region.