ScholarWorks@숙명여자대학교

초록

As sensing devices and simulation programs are widely used, a large amount of output data is being generated in the form of a 2-dimensional (2D) array. To facilitate the post data processing, it is of critical importance to find anomalous or outlying elements in that array with little human intervention. In this paper, we propose an effective method for locating outlying regions in a 2D array, in which a group of adjacent elements in its entirety deviate significantly from the entire array. To find such outlying regions, we divide the array into small subarrays and build a regression model for each subarray. We then cluster subarrays that are adjacent to each other and have similar regression models into larger subarrays. After the clustering, we detect relatively small clusters as outlying regions. Our experiments confirm the effectiveness of the proposed method. © Springer Nature Singapore Pte Ltd. 2019.

키워드