A Data Science Technique for Cropping Localization From the Weather Dataset

Abstract

Bangladesh is an agricultural country, and it is the backbone of our nation. Agriculture employs over half of Bangladesh's people, and crops occupy more than 70% of the country's territory. More than 12 percent of revenue comes from the agriculture sector. But our lands are limited, and our population is growing day by day. That's why we need more food and increase the demand for crops continuously. So, it is a huge challenge to increase crop production. But our farmers face different types of problems during cultivation. They cannot justify which crop should be cultivated. Because of this, they did not get the expected yield. We know that machine learning plays a vital role in agricultural prediction. Crop prediction is a complex process. A massive amount of data is needed, like temperature, humidity, precipitation, wind speed, dew etc. In this system, we applied different types of machine learning algorithms and checked which algorithm gives us better accuracy. We get Random Forest gives us the best accuracy. So, we applied it to our dataset. We collect weather data from NASA Power Access Viewer. And crop data from different sources. Then we apply training and testing to this dataset. We took 80% data for training and 20% for testing. We get 91% accuracy from the Random Forest algorithm, which will help the farmer to decide which crop should be cultivated. It will increase crop production. It Removes hunger and poverty from our country.