a wafer metrology consortium under MAGNET

Metro450 Members Publications

Lookahead Selective Sampling for Incomplete Data

28 June, 2016

Lookahead Selective Sampling for Incomplete Data


AU: Loai AbdAllah and Ilan Shimshoni

SO: International Journal of Applied Mathematics and Computer Science, 26(4)



Missing values in data are common in real world applications. There are several methods that deal with this problem. In this paper we present lookahead selective sampling algorithms (LSS) for datasets with missing values. We developed two versions of selective sampling. The first one integrates a distance function that can measure the similarity between pairs of incomplete points within the framework of the LSS algorithm. The second algorithm uses ensemble clustering in order to represent the data in a cluster matrix without missing values and then run the LSS algorithm based on ensemble clustering instance space (LSS − EC). To construct the cluster matrix we use the kmeans and the mean shift clustering algorithms especially modified to deal with incomplete datasets. We experimented our suggested algorithms on six standard numerical datasets from different fields. On these datasets we simulated missing values and compared the performance of the LSS and LSS − EC algorithms for incomplete data to other two basic methods. Our experiments show that the suggested selective sampling algorithms outperform the other methods.