A Study on Improved Productivity of Crops Sown Using Machine Learning Algorithms

Abstract

The main mission of this work is to enhance agricultural productivity by guiding farmers in selecting the most-suited crop yields based on location and soil parameters. Given the widespread underutilization of technology in farming, many farmers inadvertently choose inappropriate crops, leading to reduced yields and profitability. To address this issue, the proposed system leverages machine learning (ML) algorithms for the prediction of the best-suited crops for specific soil conditions, thereby increasing productivity and profitability. The study employed four ML algorithms, such as support vector machine (SVM), Naive Bayes (NB), decision tree, and random forest, to analyze a dataset of soil parameters. After a comprehensive qualitative analysis, SVM emerged as the most accurate algorithm, with an accuracy rate of approximately 93%, compared to 83% for Naive Bayes, 62% for decision tree, and 72% for random forest. Whereas, the high accuracy of the SVM model ensures that the crop recommendations are reliable and suitable for the given soil conditions, minimizing the risk of undernourishment. The research novelty lies in its application of the four machine learning techniques mentioned in this paper to the agricultural domain, providing a practical tool that can significantly improve crop selection decisions. By integrating soil testing data with predictive modeling, the system offers a novel approach to optimizing crop yields, ultimately contributing to the overall productivity of the farming sector and the nation.