Artificial Intelligence based Evolving Ensemble Learning Model for Evolving Data Stream Classification

Data Stream classification is an emerging area in current data mining research. In recent years, Artificial intelligence technique has been employed for large data stream classification on exploration of feature extraction and knowledge representation process to categorize the data streams into one or more classes. Unsupervised data classification based on ensemble Artificial intelligence has been employed in this work to predict the classes for the outlier data in the data streams which is considered as imperfect labels from training samples for all features in the data stream on analysis. In this paper, a novel artificial intelligence technique combination has been presented for data stream classification. It is named as Evolving Ensemble learning model as a multistep learning process utilizing the infrequent principle Component Analysis, K Nearest Neighbour and Expectation maximization Algorithm of the artificial intelligence technique to generate the new classes from its regularized classes with data outliers. In addition, Markov hidden model has been used in addition to detect the latent feature in the data stream to construct the feature set. It is capable of detecting the feature and concept of evolution on the feature space and label space of the classes. Further feature reduction technique has been employed to remove or eliminate the curse of dimensionality and sparsity issue in terms irrelevant, Noisy and redundant features and reduces the dimensionality of feature space. Experiment results explain the effectiveness of the proposed Al based ensemble model against the state of art approaches in large stream data classification. Proposed classification model outperforms the existing model on performance metrics like precision, recall, F measure and Accuracy and classification error.