Hellinger Divergencive Theil–Sen Regression-Based Deep Multilayer Perceptive Feed Forward Network for Predictive Analytics with Big Data

Abstract

Predictive analytics was the technique with an advanced research approach to find future events, especially in the agriculture field. Agriculture of precision is the leading part within the development of the financial system. Climate as well as soil quality conditions changes have become the most important risk in the prediction of crop yield in agriculture field. But, an accurate prediction faces high complexity. Hellinger divergence matching pursuit Jaspen's correlative regression-based deep multilayer perceptive feed-forward Network (HDMPJCR-DMPFFN) is introduced for increasing accuracy by lesser complexity. Deep feed forward Network includes several layers for processing the given input data. Deep multilayer perceptive network has input layer that collects a large volume of data. Next, input was transformed of deep neural network where the feature selection was performed by Hellinger divergence target matching pursuit. By applying Hellinger divergence, the significant target features related to the prediction are selected. The feature selection helps to reduce complexity time as well as space complexity. Next, selected features were given to subsequent thidden layer to classify the information based on the Multivariate Jaspen's correlative Theil–Sen regression. The regression function detects training features as well as testing features by Multivariate Jaspen's correlation to classify the data for accurate prediction with higher accuracy. Experimental evaluation of HDMPJCR-DMPFFN has various parameters like prediction accuracy, false-positive rate, prediction time, and space complexity. HDMPJCR-DMPFFN enhances accuracy of prediction and decreases the time consumption and space complexity of prediction methods.