10th International Congress on Information and Communication Technology

Authors - A S M Ahsanul Sarkar Akib, Abu Zahid Md Jalal Uddin, Fatema Jahan Sifa, Mahadir Islam, Md. Easin Arafat, Touhid Bhuiyan
Abstract - This paper presents an innovative IoT-based autonomous farming system utilizing machine learning models to assist farmers in determining suitable crops based on real-time environmental data. The system integrates IoT sensors, including soil pH, NPK, temperature, and humidity sensors, to collect data from the field. The ESP-8266 NodeMCU processes this data and transmits it to a cloud database. A range of machine learning algorithms were applied to the dataset, including Logistic Regression, Gaussian Naive Bayes, Support Vector Classifier, K-Nearest Neighbors, Decision Tree Classifier, Extra Trees Classifier, Random Forest, Bagging Classifier, Gradient Boosting, and AdaBoost. The highest accuracy was achieved with the Random Forest Classifier (97.05%), followed closely by the Bagging Classifier (96.59%) and Gradient Boosting (96.36%). The AdaBoost model showed poor performance with an accuracy of 10.23%. The system’s predictions are accessible to farmers via a web or mobile application, enabling them to make informed decisions about crop cultivation. This IoT and machine learning-based approach reduces human intervention, optimizes farming practices, and enhances crop yield potential. The system provides real-time crop recommendations, making farming more efficient and sustainable. Use of appropriate algorithms on the sensed data can help in recommendation of suitable crop.