Machine learning and data science for decision making in crop cultivation

Abstract

In many situations, farmers face lots of difficulties due to not gaining enough income for their harvest as a result of it, considerable amount of harvest is wasted because of not having enough return on investment. Although government introduced many approaches, they seem to have been unable to address the issues of crop prediction and sustainable agriculture as farmers complain that the policies created by the authorities never match with the expectations of farmers’ community. Considerable amount of population still engage in agricultural sector. Therefore, their contribution to the GDP has been representing an important portion through last few decades. Therefore this research has been conducted to support farmers to make optimum decisions based on various factors such as weather, price, and potential production to make cultivations. The research is based on two basic data mining models (Price and Expected Production) to predict the price and Production for a selected crop based on a particular district. The prediction outcome is presented using a mobile application with user friendly interfaces. Price model is based on the production (MT), per capita consumption, imports and inflation with the accuracy of 77.93% and the Production model is based on the climatic conditions such as humidity, wind speed, sun hours, rainfalls and temperature with the accuracy of 65.08%. The production model is to be further developed by adding the feature of soil moisture as it is an important factor to decide which crop is to be cultivated. Furthermore, an additional model is to be added to predict the amount of fertilizer to be added based to get the optimum yield.