IOT ENABLED DISCRIMINANT FUNCTION ANALYSIS AND DEEP MAP REDUCE CLASSIFICATION FOR USER BEHAVIOUR PREDICTION WITH BIG DATA

Web usage mining is type of web mining activity with user access pattern discovery from several web servers. Many researchers carried out their research on frequent pattern mining for user behavior prediction. But, time consumption and space complexity were not minimized by existing methods. IoT enabled Discriminant Function Analyzed Deep Multi-Layer Perceptive Map Reduce Classifier (IoTDFADMLPMRC) Model is introduced to improve accuracy and to minimize time consumption during web user behavior analysis. IoT-DFADMLPMRC Model comprised one input layer, two hidden layers, and one output layer. Weblog files are collected using IoT devices and transmitted to hidden layer 1. In that layer, Discriminant Function Analyzed Preprocessing (DFAP) identifies linear combination of patterns that separate into relevant and irrelevant web patterns. After that, irrelevant web patterns are removed and relevant web patterns are transmitted to hidden layer 2. In that layer, Zijbendos Similarity Coefficient is used for finding frequently accessed web patterns. Finally, in output layer, user behavior gets analyzed based on similarity score with higher accuracy. Experimental evaluation of IoTDFADMLPMRC Model is conducted with weblog dataset from Kaggle on factors like prediction accuracy, error rate, prediction time, and space complexity with respect to number of weblogs.