Data Streams Classification Method Handling Concept Drifts
Xiaoye Wang · Journal of Information and Computational Science · 2014
This paper presented a classification algorithm which could find the concept drifts from the data streams, called Incremental Updated Discriminant Eigenspace (IUDE). The algorithm constructs the eigenspace model of the data streams based on the algorithm of LDA (linear discriminant analysis). It handles the Gradual Concept Drifts by incremental update the Discriminant eigenspace. And at the same times, it uses the Mean Square Error (MSE) to handle the Abrupt Concept Drifts. The 1-nearest neighbour (1-NN) method classifies the new data streams. The experiment result shows that it is better flexibility over concept drifts than traditional approaches.