Implementing Machine Learning Models – An Analysis of Agricultural Weather And Soil Data

Abstract

Agricultural data based on various characteristics such as water, soil type, weather, and environment with regional and geographical variations is very important for smart agriculture systems. Automated irrigation, precision planting, variable rate application in plant nutrition, temperature and humidity prediction, water quantity prediction are most common in these day. Computer vision and artificial intelligence has introduced new trends in precision agriculture production. Internet of Things (IoT) sensing devices can sense temperature and steam in the mining area where most of the critical condition is faced on regular basis. This article analyses the crop yield prediction from statistical data obtained by deploying sensors in the crop fields. The area of the crop covered is about two acres i-e 87120 Sq. foot. The dataset is constructed on the basis of four important parameters of weather and soil including temperature, soil pH level, humidity, soil moisture and smoke. An IoT based kit having DHT, soil moisture and D11 sensors connected with Adriano Uno was used to collect the data. The two major crops monitored named Wheat and Cotton for dataset construction. The data set attributes are selected by applying principal component analysis. The highest ranked attribute is selected for data analysis. ML models parameters are Correlation coefficient, Mean Absolute error, Root Mean Squared Error, relative absolute error and Root relative squared error. The results are obtained by Applying WEKA Data Mining Tool. The results showed that Random Forest Classifier performed better results as compared to KNN and Decision Tree classifiers.