A machine learning framework for profit analysis: A case study on sustainable cocoa production in Ecuador

초록

The transition to organic agriculture is recognized for sustainability benefits, yet adoption remains low due to economic uncertainties. Cocoa farming in Ecuador faces challenges balancing profitability and sustainability, making financial analysis crucial.This study analyzes the profitability of organic versus conventional cocoa farming in Ecuador using machine learning models to identify key economic and environmental indicators.We analyze national agricultural survey data covering 10,605 farm-year records (2020–2022). We compare several prediction methods, ranging from simple ones (linear regression, nearest-neighbors) to more advanced tree-based and neural-network methods (Random Forest, gradient-boosted trees such as XGBoost and HistGradientBoosting, and a Multi-layer Perceptron). We then use an explainability technique (SHAP) to rank which factors most influence per-hectare profit. The factors examined include the farmer’s perceived quality of life, planted area, soil quality, fertilizer use, plantation age, irrigation, and the market price of cocoa.Results indicate organic farming outperforms conventional Fine Aroma cocoa in profitability (measured as gross margin over agronomic inputs) for the Fine Aroma variety, while conventional farming is more profitable for the Ramilla variety. When all factors are considered together, the farmer’s perceived quality of life is the strongest single predictor of profit, and the models explain profit well (the share of variation explained, R², is 0.80–0.87). However, when quality of life is removed, predictive accuracy falls sharply (R² of 0.06–0.11). This shows that quality of life is best read as a summary signal that moves together with profit, not as a lever that can be acted on directly. Once it is set aside, the factors that can actually be managed come into focus: plantation age, irrigation, cocoa market price, chemical fertilizer use, and farm area. For policymakers and businesses, the practical implication is that support should target these manageable factors. This includes renovating aging plantations, investing in irrigation, and stabilizing prices through instruments such as forward contracts and cooperative bargaining, supported by better market monitoring and forecasting, rather than attempting to influence perceived quality of life directly. © 2026 Elsevier B.V. All rights are reserved, including those for text and data mining, AI training, and similar technologies.

키워드