Plant diseases are one of the primary brakes to agricultural productivity, as they both lower crop yield and quality. The use of ML techniques allows diseases to be identified and diagnosed through image analysis. Predictive analytics can also be used to determine the breakout of disease, saving resources through precision agriculture. In this work, we propose to compare CNN, AlexNet, and InceptionV3 for detection of plant disease based on the leaf images. The dataset involved in the paper is obtained from Kaggle, and all the models are tested rigorously based mainly out of accuracy as primary metric. Apart from disease detection, the paper postulates an integrated system with the capability to not only detect plant diseases but also provide practical recommendations for fertilizer application along with precautions for the disease’s mitigation. In this study, the ExtraTrees Classifier achieved 88% accuracy for predicting fertilizer types, while the XGBoost Regressor delivered the best performance in fertilizer quantity prediction with a Mean Absolute Error (MAE) of 3.32. These models offer effective solutions for optimizing fertilizer use in agriculture. The study attempts to empower agricultural practitioners with tools for informed decision-making by integrating the techniques of machine learning with actionable insights. Our research is a critical bridge that integrates comparative analysis of the different architectures in neural networks with a holistic approach to disease management in agriculture. In so doing, we incorporated practical recommendations into our study, and our researches go beyond theoretical analysis to directly influence agricultural practices, thereby improving crop health and productivity. Promising results were also obtained, showing classification accuracy from 97 to 100% across classes The proposed model records an with nine classes of diseases and one class of a healthy plant, the proposed model achieved average accuracy of 99.21%. As such, these findings provide knowledge to indicate the performance. its application to disease diagnosis and classification in plants and also improving the feasibility of our approach in the field of agriculture.