A Review on Limited Labeled Data in Data Stream

Authors

  • Hiral Desai Computer Engineering Department, B.H.Gardi College Of Engineering and Technology
  • Jay Gandhi Computer Engineering Department, B.H.Gardi College Of Engineering and Technology

Keywords:

Data stream mining; Classification; Semi-Supervised Learning; Limited Labeled Data; Clustering

Abstract

Assigning labels to unlabeled data in data streams from the few labeled data is a momentous and
interesting issue for machine learning and data mining that commonly looks in real world stream classification
hitches. Till now there are few research work has been done to labeled the unlabeled data for stream data, but Most
existing work on classification of data streams take that all streaming data are labeled and the class labels are
promptly available. However, in real-life uses, such as fraud and intrusion detection, this suspicion is not generally
valid. It has more costly and time consuming to labeled the data manually. Accuracy of labeling the data stream is
between 34% to 95% . But for image dataset of stream data we have maximum accuracy is 78%. In proposed
approach with the use of SVM (support vector machine) and co-relation to the image type of stream data we can
improve accuracy for labeling.

Published

2016-12-31

How to Cite

Hiral Desai, & Jay Gandhi. (2016). A Review on Limited Labeled Data in Data Stream. International Journal of Advance Research in Engineering, Science & Technology, 3(13), -. Retrieved from http://ijarest.com/index.php/ijarest/article/view/2167