An accurate prediction of crop yield using hybrid deep capsule auto encoder with softmax regression

Crop yield prediction is every nation's most important and difficult task. Due to the frequent climate changes, farmers struggle to achieve a high yield. This work presents an effective methodology for crop yield prediction of major crop production. In the presented methodology, crop yield is predicted by considering soil health data, crop production data and rainfall data. The input soil nutrients, weather and crop production data are processed. The input data is collected from the Uttar Pradesh (UP) region. In the first phase, the soil fertility index estimates the soil quality. In the second phase, the soil quality score and other crop yield related parameters like rainfall and crop production data are taken for further processing. Initially, the input soil health, crop production, and rainfall data are pre-processed. In the pre-processing stage, data cleaning with if-null processing and min-max normalization is used for data standardization. The features are then selected using the adaptive bald eagle search optimization (ABES) approach. Finally, the crop yield prediction of sugarcane, wheat and rice crops is obtained accurately by utilizing a hybrid deep capsule auto encoder with a softmax regression (Hybrid DCAS) model. Here, the hyper-parameter tuning of the presented deep learning model is achieved by a modified Flamingo Search (MFS) optimization approach. The overall implementation is carried out on PYTHON. The performance of the developed model is compared with other models in terms of classification and prediction performance.

Automated summary

Crop yield prediction is important and difficult. This study presents an effective methodology that considers soil health, crop production, and rainfall data for accurate prediction.