Explainable AI Models for Interpretable Lettuce Prediction in Indoor Agriculture

About the Project

Indoor agriculture is a sustainable way to produce food. But the majority of machine learning models are not easily explainable, which hinders the understanding and confidence of users in their predictions. The goal of this project is to create an Explainable Artificial Intelligence (XAI) system for estimating the weight of lettuce shoots that can be used for consumption, by leveraging environmental and nutrient factors. 216 lettuce harvest data combined with eight environmental data were used to train and test three machine learning algorithms: Support Vector Regression (SVR), Random Forest and XGBoost. The performances of the model were evaluated by Mean Absolute Error (MAE), Root Mean Squared Error (RMSE) and R-square. Results showed that SVR had the best overall performance with an MAE of 22.94 g, RMSE of 30.75 g, and R² of 0.609. To understand model predictions and the key environmental factors, SHAP and LIME were added. An interactive Streamlit dashboard was created to offer lettuce weight predictions, model comparisons, explainability features and what-if scenario analysis. In addition, a harvest efficiency module was included to evaluate biomass usable and suggested quality adjusted yield indicators. The innovation will facilitate transparent and data-driven decision making in the context of indoor agriculture and raise awareness of sustainable food production and the use of resources. The project supports Sustainable Development Goals (SDGs) 2, 9, 11 and 12, which help drive smart and sustainable agriculture.