Smart Agricultural System Using Machine Learning to Predict Rainfall and Conserve Water for Irrigation

The primary objective of this research is to conserve water through smart agriculture, by employing the Random Forest algorithm to predict rainfall. By avoiding unnecessary irrigation on rainy days, this system conserves water. The proposed method works on both hardware and software. The hardware comprises two boards, one of which is the ESP8266 for wireless connectivity to the internet, while the other utilizes an Arduino microcontroller with additional sensors to track soil moisture, temperature, and humidity of the farming land. The software is a website developed using the Random Forest algorithm for rain prediction. The project integrates hardware and software by providing the sensor outputs through the website to the machine learning model for rain prediction. If rain is predicted, land water utilized for farming is conserved. With an accuracy of 88.5%, Random Forest predicts rain on that day, whereas the other models give an accuracy of 83%. The Random Forest algorithm also outperforms logistic regression used for rain prediction, resulting in more accurate predictions.