AN ACCURATE CLASSIFICATION OF IMBALANCED STREAMING DATA USING DEEP CONVOLUTIONAL NEURAL NETWORK

The problem of class imbalance is irritated when learning from data streams, as the duration between consecutive positive class examples can become arbitrarily large, which in turn may seriously impair the learner's ability to learn the positive class. Ensemble-based methods are among the most widely used techniques for data stream classification. Their popularity is attributable to their good performance in comparison to strong single learners while being relatively easy to deploy in realworld applications. Ensemble algorithms are especially useful for data stream learning as they can be integrated with drift detection algorithms and incorporate dynamic updates, such as selective removal or addition of classifiers. We propose an ensemble based Deep CNN (Convolutional Neural Network) algorithm for streaming imbalanced data classification. The proposed model is designed to handle and classify high dimensional data. So the dimensionality is reduced with PCA (Principal Component Analysis) method. The redundant and irrelevant attributes are eliminated by the heuristic based feature selection algorithm. This is done to improve the accuracy of classifying the instances. The analysis includes the elimination of irrelevant and duplicate data so as to improve the accuracy to further extent. Learning will help to classify the instances more accurately and refine the boundaries of the classes. We conduct experiments on real-world data sets which are taken from MOA (Massive Online Analysis) and UCI machine learning repository. The predictive ability of classifiers usually evaluated in terms of Accuracy, G-mean, Mean square error, Recall of minority and sensitivity