Development of a machine learning algorithm for soil moisture prediction:

Abstract

This study investigated soil moisture content with real data measured from cropped farm areas especially irrigation schemes. Soil moisture content is important in irrigation scheduling (when to irrigate and how much to irrigate), which is key in the growth and development of the agricultural sector. In order to aid farmers, get real time soil moisture predictions, we have designed a locally hosted web application using a maize farmland dataset obtained from Kaggle.com. the application can be used to predict soil moisture to a tune of 80% accuracy. This application can be used round the clock. In building the model, 5 machine learning techniques were used and these were Random Forest Algorithm, Gradient Boosting regressor, the lasso regressor, elastic net regressor and the ridge regressor. With this application, farmers can obtain soil moisture readings without the need for soil moisture probes.