A Comparison among Data Mining Algorithms for Outlier Detection using Flow Pattern Experiments

Abstract

Abstract. Accurate outlier detection is an important matter to consider prior to applying data to predict ow patterns. Identifying these outliers and reducing their impact on measurements could be e ective in presenting an authentic ow pattern. This paper aims to detect outliers in ow pattern experiments along a 180-degree sharp bend channel with and without a T-shaped spur dike. Velocity components have been collected using 3D velocimeter called Vectrino in order to determine the ow pattern. Some of outlier detection methods were employed in the paper, such as Z-score test, sum of sine curve tting, Mahalanobis distance, hierarchical clustering, LSC-mine, self-organizing map, fuzzy C-means clustering, and voting. Considering the experiments carried out, the methods were e cient in outlier detection; however, the voting method appeared to be the most e cient one. Brie y, this paper calculated di erent hydraulic parameters in the sharp bend and made a comparison between them for the sake of studying how e ective running the voting method is in mean and turbulent ow pattern variations. The results indicated that developing the voting method in the ow pattern experiment in the bend would cause a decrease in Reynolds shear stress by 36%, while the mean velocities were not signi cantly in uenced by the method.

Forecasting natural gas production and consumption using grey model with latent information function: The cases of China and USA

“…Larger e(k) with larger LIe(k) and smaller e(k) with smaller LIe(k) are not logical. These unreasonable errors are caused by outliers, therefore these outliers must be corrected in order to improve prediction accuracy [27].…”