An extensive three-tiered architecture for comprehensive crop and fertiliser prediction using supervised learning

Agriculture accounts for a fifth of India's GDP, however the research in this field does not reflect this significant contribution. Farming practices remain archaic, with little emphasis on data-driven approaches to maximising yield and profits. Predicting crop yield is crucial for maximising profits in agronomy, with suitable fertiliser selection vital for maintaining soil health. This paper presents an extensive three-tiered architecture for comprehensive crop and fertiliser prediction using historical data with features such as soil pH, moisture, and temperature. The first tier predicts crops based on the area under cultivation and geographical region, with an accuracy of 99.54% using the random forest classifier. The yield for the given crop is predicted using linear regression with an accuracy of 89.57%. The second tier predicts the cost of cultivation, and the third predicts an appropriate fertiliser based on soil nutrients and environmental factors using Naïve Bayes with 100% accuracy.