A Novel Design and Implementation of Fertilizer Recommendation System Based on Hybrid Machine Learning Models

Abstract

India is mostly a farming country. Many of the GDP of emerging nations like India comes from agriculture, so the sector is important to these countries' economies. The demand for food has skyrocketed due to the increase in population. Crop quality, yield, and profitability can all take a hit when farmers choose their crops, fertilizers, and pesticides without considering factors like soil type, water requirement, temperature conditions, and crop profitability analysis for a specific area. The development of computational technology has prompted scientists to consider various issues, including identifying diseases and pesticides and selecting fertilizers and crops based on soil quality, water needs, and market viability. An essential and fundamental aspect of farming is selecting the appropriate fertilizer for soil and crop production. India has a reputation as an agricultural powerhouse, with traditional practices still used to advise farmers on the best fertilizer to use. Communication between farmers and specialists is currently the basis for suggestions, and various experts have different recommendations. The prohibitive cost of lab technology for determining soil supplement levels is a major concern. Present frameworks for determining soil nutrient content and fertilizer recommendations are ineffective and inefficient. In order to estimate the nutritional dimension in soil and provide suitable fertilizer, this article presents an attractive, novel fertilizer recommendation system, 'FertRec'. In this proposed system, soil samples are analyzed to identify the deficiency of nutrients, thus preparing the datasets used for training the machine learning models. The best accuracy model recommends a suitable fertilizer based on the soil nutrients. The main goal is to create an effective fertilizer recommendation system to help farmers optimize their fertilizer use. Compared to current benchmark recommendation methods, the proposed system performs four times better on 500 soil samples from the Telangana region in India, using soil features, and it effectively recommends fertilizer with 99.98% accuracy. This will maximize production, and farmers greatly benefit from this method, which involves selecting the appropriate fertilizer at the beginning of the product cycle.