Machine Learning Based Crop Recommendation System

Agriculture is very important in the Indian economy. Nowadays, due to the change in climate and the increase in global warming, the weather is an unpredictable variable. So, the most common issue that Indian farmers encounter is that they fail to identify the best-suited and appropriate crop for their soil using conventional methods. As a result, they experience a significant drop in production. This is a big problem in a country where farming employs over 58 percent of the population and results in low crop production. To overcome this issue, a model is built using machine learning which has a better system to guide the farmers, and it is a modern agricultural strategy for selecting the best crop by considering all the factors like nitrogen, phosphorus, potassium percentages, temperature, humidity, rainfall, and ph value. This paper proposes the use of machine learning techniques such as logistic regression, decision tree, KNN (k-Nearest Neighbours) and Naive Bayes to determine the best-suited crop based on attributes of soil and environmental factors. In the end, an accuracy of 96.36 percent from the logistic regression, 99.54 percent from the decision tree, 98.03 percent from the k-nearest neighbours and 99.09 percent from the naive Bayes is obtained, resulting in the decision tree having the highest accuracy with 99.54 percent. This paper gives an extensive Exploratory Data Analysis (EDA) on the Crop recommendation Dataset and builds an appropriate Machine Learning Model that will help farmers predict their suitable crops based on their parameters.

Keywords: Accuracy, Crop recommendation, Dataset, Data pre-processing, Decision tree, Humidity, k-nearest neighbours (KNN) algorithm, Logistic Regression, Machine learning (ML), ML algorithms, Naive Bayes algorithm, Ph value, Python, Rainfall, SciKit-learn, Soil NPK percentages, Streamlit, Temperature.