Incremental Data Stream Classification with Adaptive Multi-Task Multi-View Learning

With the enhancement of data collection capabilities,massive streaming data have been accumulated in numerous application scenarios.Specifically,the issue of classifying data streams based on mobile sensors can be formalized as a multi-task multi-view learning problem with a specific task comprising multiple views with shared features collected from multiple sensors.Existing incremental learning methods are often single-task single-view,which cannot learn shared representations between relevant tasks and views.An adaptive multi-task multi-view incremental learning framework for data stream classification called MTMVIS is proposed to address the above challenges,utilizing the idea of multi-task multi-view learning.Specifically,the attention mechanism is first used to align different sensor data of different views.In addition,MTMVIS uses adaptive Fisher regularization from the perspective of multi-task multi-view learning to overcome catastrophic forgetting in incremental learning.Results reveal that the proposed framework outperforms state-of-the-art methods based on the experiments on two different datasets with other baselines.

Exploiting Empirical Variance for Data Stream Classification

Classification,using the decision tree algorithm,is a widely studied problem in data streams.The challenge is when to split a decision node into multiple leaves.Concentration inequalities,that exploit variance information such as Bernstein's and Bennett's inequalities,are often substantially strict as compared with Hoeffding's bound which disregards variance.Many machine learning algorithms for stream classification such as very fast decision tree(VFDT) learner,AdaBoost and support vector machines(SVMs),use the Hoeffding's bound as a performance guarantee.In this paper,we propose a new algorithm based on the recently proposed empirical Bernstein's bound to achieve a better probabilistic bound on the accuracy of the decision tree.Experimental results on four synthetic and two real world data sets demonstrate the performance gain of our proposed technique.

Clustering feature decision trees for semi-supervised classification from high-speed data streams

Most stream data classification algorithms apply the supervised learning strategy which requires massive labeled data.Such approaches are impractical since labeled data are usually hard to obtain in reality.In this paper,we build a clustering feature decision tree model,CFDT,from data streams having both unlabeled and a small number of labeled examples.CFDT applies a micro-clustering algorithm that scans the data only once to provide the statistical summaries of the data for incremental decision tree induction.Micro-clusters also serve as classifiers in tree leaves to improve classification accuracy and reinforce the any-time property.Our experiments on synthetic and real-world datasets show that CFDT is highly scalable for data streams while gener-ating high classification accuracy with high speed.