Improved feature ranking fusion process with Hybrid model for crop yield prediction

Former and accurate prediction of the crop yield is vital for statistical as well as economic valuation on the farm levels to govern investment plans in agricultural products. Among the difficult problems in the agricultural industry, crop yield predictions performed to forecast the higher yield of crops using artificial intelligence techniques faces more complications. Considering the increasing significance of crop yield predictions, this paper proposes a new crop yield prediction model by utilizing hybrid classification model based on the improved feature ranking fusion process. In this model, initially the unnecessary data is cleansed by Data Normalization and subsequentl By an improved SMOTE algorithm is proposed that enhances the data to make it proper for feature extraction. The data features are essential to analyses the respective in-depth information, hence, the feature extraction process includes the extraction of Improved Correlation based features, Statistical features, Entropy features and Raw Data. In order to ensure the selection of most important features, it is necessary to make optimal feature selection. Therefore, an improved feature ranking fusion process is employed to choose the suitable features, in which, the results of three feature selection methods like chi-square, Relief and RFE are included. Finally, the prediction process is carried out by the proposed hybrid model, which is the combination of LSTM and DBN models. Finally, the performance of proposed work is validated in terms of accuracy, precision, specificity, sensitivity and then the results show that compared to the conventional classifiers such as LSTM, DBN, CNN, Bi-GRU, SVM, respectively.