Unsupervised Double local weighting for feature selection
Nadia Mesghouni, Moncef Temanni · 2011
In this paper we proposed a new method Double local weighting based in self organized map (som), features weighting and on two learning methods local-observation-Som and local-distance-Som. This method allows us to weight the observation and the distance simultaneously and avoid the user to choose the confidence criteria for the weighted approach observation or distance during the learning process. We illustrate the performance of the proposed method using different data, showing a better performance for new algorithm. We can also show that through deferent means of visualization, DIS-SOM, OBS-SOM, and Dlw-SOM algorithms provide various pieces of information that could be used in practical applications.