An Efficient Classification Approach for Novel Class Detection by Evolving Feature Datastreams
Journal Title: INTERNATIONAL JOURNAL OF COMPUTER TRENDS & TECHNOLOGY - Year 2013, Vol 6, Issue 3
Abstract
Data stream classification has been an extensively studied research problem in recent years. data stream classification requires efficient and effective techniques that are significantly different from static data classification techniques because of its dynamic nature. Existing system faces major challenges in the methods namely feature-evolution, infinite length, concept-drift and concept-evolution. To address and overcome the problems in these techniques an ensemble classification framework is proposed where each classifier is equipped with a novel class detector and addresses concept-driftand concept-evolution. It also addresses feature-evolution by a technique called a feature set homogenization technique. It also enhances the novel class detection module by making it more adaptive to the evolving stream. And make this enable to detect more than one novel class at a time. But all of methods doesn’t support for the detection of the outlier class by using clustering methods. To overcome this problem Outlier Detection has been proposed which is a very important research problem in data mining. These outliers are detected efficiently by using clustering algorithms. CLARANS clustering algorithm is proposed for detecting outliers in the class. The outlier class is detected before the novel class detection algorithm is performed. The best outlier in the class can be found and then it is applied to MCM (multiclass miner) in data streams. It is more adaptive technique to the evolving stream and enabling it to detect more than one novel class at a time. Comparison with state-of-the-art data stream classification techniques establishes the effectiveness of the proposed approach.
Authors and Affiliations
R. Shree alaguvidhya , C. Yamini
Keywords
Related Articles
- EP ID EP146842
- DOI -
- Views 127
- Downloads 0
An Effective Method of Eye Detection Using Segmentation and Projection
Face detection and eye detection has long been the subject of interest in practical implementation of image processing and vision. Moreover the exact pupil detection of eye is the cream of research as the basic step of r...
ACO, Its Modification and Variants
Ant colony optimization (ACO) is a P based metaheuristic algorithm which has been proven as a successful technique and applied to a number of combinatorial optimization problems and is also applied to the Traveling sales...
Identity Management issues in Cloud Computing
Cloud computing is providing a low cost on demand services to the users, omnipresent network, large storage capacity due to these features of cloud computing web applications are moving towards the cloud and due to this...
Improving the Efficiency of Call Admission Control in Wireless Cellular Communication Networks by Frequency Sharing Techniques
In this paper, a two-tier cellular communication wireless network is characterized by overlapping the service area for managing the new calls users having different mobility speed. The overlapping property of the two-tie...
Performance Estimation of 2*3 MIMO-MC-CDMA in Rayleigh Fading Channel
In this paper we analyze the performance of 2*3 MIMO-MC-CDMA system in MATLAB which greatly reduces BER by increasing the efficiency of system. MIMO and MC-CDMA system arrangement is used to decrease bit error rate and a...