초록

용어보기논문에서 용어와 풀이말을 자동 추출한 결과로, 시범 서비스 중입니다. AI-Helper

Abstract ▼ AI-Helper

In data stream environments classifiers have to be refined because distributions that are different from the current data distribution can occur. However, we can not know if the current datum is driven from the current data distribution or not. According to the reason, classifiers are generally refi...

In data stream environments classifiers have to be refined because distributions that are different from the current data distribution can occur. However, we can not know if the current datum is driven from the current data distribution or not. According to the reason, classifiers are generally refined based on regular time interval. In such a approach if a period of time is predefined as a small value, a classifier that maintains the current performance can be unnecessarily refined by human experts. If it is predefined as a large value, a classifier can not be refined more quickly when its classification accuracy is decreasing. In this paper, we select data for refining a classifier from streaming data in an online process. Our selection methodology uses training data, and is applied to built an ensemble of classifiers over streaming data. We compared the results of our ensemble approach and of a conventional ensemble approach where new classifiers for an ensemble are periodically generated. In experiments with nine benchmark data sets including three real streaming data sets, our ensemble approach generated an average of 5.1% classifiers using an average of 10.2% labeled data for the conventional ensemble approach, and produced comparable classification accuracy. To exhibit our approach efficiently generates new classifiers for an ensemble over streaming data, we analyzed the three real streaming data sets in a time domain.

주제어