ABSTRACT
Recently, there are often changes in atmospheric conditions. Such, climatic changes make farmers fail to produce enough crops at the right time. Farmers miss to predict the crop yield. This results in insufficient yield throughout the year for a country. So that the country is forced to import some of the food resources from outside nations. Prediction of everyday weather conditions are closely related to every field by means of production, life, security, etc., forecasting the weather conditions plays an important role in social development. To improve overall crop productivity temperature, rainfall, soil, seed, humidity, wind speed data needs to be collected and analyzed, which may help farmers to improve their agriculture yield. For a successful harvesting, a farmer should be aware of climate that is suitable for sowing seeds, fertilizers to be used, watering seeds, types of pests that can affect yield in a particular season, expectancy of rainfall, breeze, etc. Soil moisture, temperature, and rainfall are very important factors influencing productivity of crops. Applying deep neural networks (DNN) includes different hidden layers, which helps to predict crop cultivation and yield effectively. A combination of recurrent neural network (RNN) and back propagation network (BPN) can be applied for crop prediction. Hence, DNN-based ensemble methods for crop prediction improves productivity by analyzing weather, which may seem to be scalable, simple, and inexpensive in agriculture yield over a large area in a country. With the publicly available sparse multi source data, transfer learning (TL) is applied, data is interpreted, and then estimated for enhancing crop yields.
