Predictive Models for Strawberry supply estimation in Cundinamarca. A comparative Approach Using Machine Learning and Regression

Abstract

Strawberry cultivation in Cundinamarca represents a key agricultural activity with a significant regional economic impact. This study aims to evaluate and compare different forecasting models to estimate the future supply of strawberries, considering variables such as cultivated area, yield, and production. Time series methods (Holt and Brown exponential smoothing), the AdaBoost algorithm, and a multiple linear regression model were applied, using historical data from the Ministry of Agriculture and Agronet. The results indicate that the multiple linear regression model exhibited the best performance, achieving a coefficient of determination of 0.9947 and outperforming the time series methods, which presented average errors exceeding 17%. Normality and homoscedasticity tests statistically validated the robustness of models. It is concluded that the proposed model provides an effective tool for agricultural planning in Cundinamarca allowing for projections of production increases associated with the expansion of cultivated areas and improvements in yield per hectare. Future research is recommended to integrate climatic variables and agricultural policy factors to enhance forecasting accuracy.