Cloud-Based Agricultural Framework for Soil Classification and Crop Yield Prediction as a Service

Abstract

Agriculture is one of the important occupations in India. Digitization in the field of Indian agriculture is in the initial stage. Indian farmers are suffering from various issues such as ignorance about soil parameters and inability to predict the yield of crops. Also, various agriculture-related information from the government agencies is not communicated to the farmers. To address the above-said issues, we have built a cloud-based agricultural framework which enables the Indian farmers, agricultural departments, and agro industries to extract useful agricultural information. The designed agricultural cloud framework is providing two services, i.e., soil classification as a service and crop yield prediction as a service. For soil classification, hybrid support vector machine (M-SVM) and for wheat yield prediction, customized artificial neural network (M-ANN) was developed. To store the agricultural data, we are using Amazon S3 and for deployment of the services, we have used Heroku cloud. The performance improvements in the range of 2–43%, 4–35%, and 1–11% were observed for M-SVM with respect to k-Nearest Neighbor (k-NN), Naïve Bayes (NB), and standard SVM classifiers, respectively. M-ANN performed with an improvement of 2% over standard artificial neural network (ANN) and 5% over multiple linear regression (MLR) models. We also observed that our agricultural cloud framework is able to provide reliable and accurate agricultural services.