Tomato Freshness Classification with Explainable AI
Six MQ gas sensors plus temperature and humidity are reduced to 26 statistical features per sample and classified into Pure Fresh, Good, or Stale to Spoiled. Random Forest, SVM and XGBoost pipelines are compared with 5-fold stratified cross-validation on the training set only, keeping the held-out test set unbiased; Random Forest was selected and scores 99.1% accuracy and F1 on the test set. Global feature importance and local LIME and SHAP explanations make each prediction auditable, served by a FastAPI backend and explored through an interactive Next.js dashboard.
Six MQ gas sensors plus temperature and humidity are reduced to 26 statistical features per sample and classified into Pure Fresh, Good, or Stale to Spoiled. Random Forest, SVM and XGBoost pipelines are compared with 5-fold stratified cross-validation on the training set only, keeping the held-out test set unbiased; Random Forest was selected and scores 99.1% accuracy and F1 on the test set. Global feature importance and local LIME and SHAP explanations make each prediction auditable, served by a FastAPI backend and explored through an interactive Next.js dashboard.