AI-BASED SOLUTIONS IN AGRICULTURE: FERTILIZER PREDICTION AND TOMATO DISEASE DETECTION USING MACHINE LEARNING AND COMPUTER VISION

This research is aimed at using artificial intelligence to improve agricultural practice in Kazakhstan. It focuses on tomato leaf disease detection and fertilizer optimization. Deep learning models – including GoogleNet (InceptionV3), VGG16, ResNet50, MobileNetV2, and a custom Convolutional Neural Network (CNN)–were evaluated for disease detection. GoogleNet had the highest accuracy of 99.72%, which shows its capability to detect tomato leaf diseases. For the optimization of the fertilizer, different machine learning-based models, namely Decision Trees, K-Nearest Neighbors, CNN, Gradient Boosting Decision Tree, LogitBoost were assessed using various PCA features. The CNN model that used six PCA features achieved the best accuracy at 97.58%. This shows how good features can help in prediction. The results show using AI technologies can significantly increase the agricultural productivity and sustainability of Kazakhstan through precise detection of diseases and optimized use of resources. In the future studies, models should be deployed to the real-time system of agriculture and also should be expanded to more crops and conditions.