An integrated IoT and machine learning framework for maize yield prediction and smart agriculture in Tanzania

Maize yield prediction plays a critical role in enhancing agricultural planning and food security, particularly in countries like Tanzania where farming is highly influenced by environmental variability. This study evaluates the performance of six machine learning (ML) algorithms—Support Vector Machine (SVM), Classification and Regression Trees (CART), K-Nearest Neighbors (KNN), Linear Discriminant Analysis (LDA), Logistic Regression (LR), and Naïve Bayes (NB)—in predicting maize yield across two distinct agro-ecological zones: Kilosa and SUA farming sites. The models were assessed using accuracy, precision, recall, and F1-score. The results show that SVM achieved the highest overall accuracy (0.59), followed by CART (0.57) and KNN (0.56), while NB recorded the lowest accuracy (0.46). Site-specific analysis revealed that CART performed best in Kilosa, demonstrating strong capability in capturing non-linear relationships, whereas KNN and SVM showed better performance in SUA, indicating their adaptability to varying data patterns. The findings further highlight that model performance varies across locations due to differences in soil conditions, climate, and management practices.