Soil Analysis and its Type Prediction with Speech Enabled Output using IoT and AWS

This paper proposes a new technique which can predict soil type and gives correct information to the farmers in audio format for improvised cultivation. It collects different soil parameters such as soil temperature, moisture and nitrogen, phosphorous and potassium (NPK) values present in the soil by taking the help of different sensors and predict the soil type using Random Forest Classifier, Support Vector Machine and Linear Regression Algorithms. From the comparison of all the mentioned machine learning algorithms, it is found that Random Forest Classifier gives the best soil type prediction with least Root Mean Square Error (RMSE) value. Using AWS technique, the predicted soil type information which is given in text format is converted to audio format which is easily understand by the farmers. A serverless application has been utilized to convey the information about the soil type to the farmers for better cultivation.