Classification is an important method for predicting class labels of samples. Although attribute-values in many real-life applications may change over time, existing classification research usually assumes that attribute-values are static. In this paper, we extend the traditional classification problem to deal with time-sequential attributes whose values may change over time. Accordingly, an algorithm called MultipleMIS-SP is presented to generate all classification rules for the classifier generation. Two scoring functions are proposed to predict class labels using our classifier. Detailed experiments are also presented. The results show that the accuracy of MultipleMIS-SP is greater than the traditional classification technique C4.5 algorithm in both the synthetic datasets and the real dataset.
[1]
Richard O. Duda,et al.
Pattern classification and scene analysis
,
1974,
A Wiley-Interscience publication.
[2]
Peter Clark,et al.
The CN2 Induction Algorithm
,
1989,
Machine Learning.
[3]
Yiming Ma,et al.
Improving an Association Rule Based Classifier
,
2000,
PKDD.
[4]
J. Ross Quinlan,et al.
Induction of Decision Trees
,
1986,
Machine Learning.
[5]
Aiko M. Hormann,et al.
Programs for Machine Learning. Part I
,
1962,
Inf. Control..
[6]
Ramakrishnan Srikant,et al.
Mining Sequential Patterns: Generalizations and Performance Improvements
,
1996,
EDBT.