Behavioural Analysis and Prediction of Big Data using Classification Models

Abstract

In this research work we are proposing Behavioural Analysis and Prediction of Big Data using Classification Models and this chapter mainly emphasizes the introductory part of the proposed research work. It highlights the motivation, objectives and scope of the research work. The proposal is trying to automate the behavioural analysis on the text interactions. Twitter media is utilized to access the interactions of the user for analysis. Twitter provides number of APIs to access the data or information for research and development activities. Tweepy is a python API used to access the data from the twitter as per the search tag.Classification approach is a method of labelling the test sample based on the available training samples. Sample is a collection of features / properties of anything. Feature is any property defines the one of the dimension of the object. Extraction and selection of the features depends on the agenda of the research activity. There are several classification algorithms in the literature namely K-NN, Hidden Markov Models (HMV), Support Vector Machine (SVM), Bayesian Classifier and many more.Training the system is nothing but collecting representative sample for a group of sample to identify given test sample. Also there are various training (learning) processes based on the type of training samples. newline Supervised learning newline Unsupervised learning newline Reinforcement learning Supervised learning is a process of identifying label for a test sample by involving already labelled training sample called as dataset. Unsupervised learning is a process of labelling the test sample based on unlabelled training samples. newlineWhereas reinforcement learning is identifying the label for test sample and adding the sample to existing dataset, whereby improving the performance of the classification process.The proposed research follows the reinforcement learning approach for its dataset construction and using K-NN for classification process newline