Integration of Cloud-Sensor Based Framework for Predictive Analysis for Accelerating Knowledge Gain in Agriculture

Precision agriculture is a fast-expanding industry that uses a variety of technology to maximize agricultural productivity, increase profitability, and reduce environmental impact. Our research, we provides an innovative method for predicting crop yield by utilizing machine learning methods and AWS services. Seven essential features are collected, which includes nitrogen, phosphorus, potassium, pH, rainfall, temperature, and humidity are collected from the sensors and datasets are prepared for this study. The features are real time collected from grape farms, stored in S3 bucket. Amazon Sage Maker, a fully managed service that enables developers and data scientists to create, train, and deploy machine-learning models, is used to create specific the machine-learning model for the prediction. The Quick Sight is used to visualize the data and obtain further insights after the model was trained on data that was immediately accessed from the S3 bucket. Based on the input features, crop yield is predicted using four different methods, including Decision Tree, Naive Bayes, Support Vector Machine, and Random Forest. The study's findings showed that, with an accuracy of 98.06%, the Random Forest algorithm performed better than the other models. Pickle is to store the trained Random Forest model, which is utilized to make predictions in the future.