A Double-Window-Based Classification Algorithm for Concept Drifting Data Streams

2010 ◽  
Author(s):  
Qun Zhu ◽  
Xuegang Hu ◽  
Yuhong Zhang ◽  
Peipei Li ◽  
Xindong Wu
Keyword(s):  
Data Streams ◽  
Classification Algorithm

scholarly journals An Ensemble Classification Algorithm Based on Information Entropy for Data Streams

2019 ◽  
Vol 50 (3) ◽  
pp. 2101-2117 ◽  
Author(s):  
Junhong Wang ◽  
Shuliang Xu ◽  
Bingqian Duan ◽  
Caifeng Liu ◽  
Jiye Liang
Keyword(s):  
Information Entropy ◽  
Data Streams ◽  
Classification Algorithm ◽  
Ensemble Classification

A Kind of Improved SPRINT Classification Algorithm Adaptive to the Change of Concept Drift

2012 ◽  
Vol 198-199 ◽  
pp. 1403-1407
Author(s):  
Jia Bing Zhao ◽  
Chun Yan Fu ◽  
Mao Song Ge ◽  
Li Ming Zhang
Keyword(s):  
Data Streams ◽  
Classification Accuracy ◽  
Concept Drift ◽  
Current Situation ◽  
Classification Algorithm ◽  
Model Reconstruction

With the wide application of data streams mining, the study on data streams classification algorithm with concept drift has become an important piece of work. In light of the characteristics of data streams, this paper puts forward a kind of improved SPRINT classification algorithm adaptive to the occurrence of concept drift. It is proved by experiments that it can automatically adjust the number of training window and new sample during model reconstruction according to the current situation of concept drift and consume less sources and have higher classification accuracy.

Decision Tree Classification Algorithm within Concept Similarity

2012 ◽  
Vol 235 ◽  
pp. 9-14
Author(s):  
Chun Hua Ju ◽  
Li Li Mao
Keyword(s):  
Data Streams ◽  
Data Stream ◽  
Concept Drift ◽  
Classification Algorithm ◽  
Streaming Data ◽  
Decision Tree Classification ◽  
The Cost ◽  
Prediction Efficiency ◽  
Concept Similarity

Data stream mining has been applied in many domains, but the concept drifts of data streams bring great obstacles to data mining. Current researches about classification algorithm for streaming data with concept drift have achieved many successes, while they pay little attention to the iterancy of data streams, namely, the situation of the historical concept reappears. For this characteristic, this paper puts forward that it utilizes the classifier model of the historical concepts or high similarity concepts through calculating the concept similarity to classify and predict. In this way, we don’t need training any more. Meanwhile, it reduces the cost of update model, speeds up the classification of the rate and improves the prediction efficiency.

Sign in / Sign up

Export Citation Format

Share Document