A reformative feature selection algorithm in fall detector application
Yuqi Cai, Zhonghua Zhao · 2016
There are many fall detector applications on the Internet, all the applications aim to run more quickly and judge the status more accurate.When meeting masses of fall detector data, it needs to use lots of mathematical or geometrical features to judge the status, which is not benefit for the application.This paper is aimed to use a feature selection algorithm to calculate the most effective features, which is significant to the detect result, to reduce the cost of the feature selection process.Therefore, this paper has proposed an improved algorithm to advance the accuracy of the selection of the most important features, and then use different classify algorithm to classify with subsets after relief algorithm and reformative relief algorithm.The result shows that the reformative relief algorithm can provide a more effective subset which can reduce the feature size and improve the accuracy of classify samples.