Detecting outliers on UCI repository datasets by Adaptive Rough Fuzzy clustering method

The clustering is the most effective method to identify the outliers in the UCI Repository dataset. This paper proposes detecting outliers on UCI datasets using Adaptive Rough Fuzzy C-Means clustering algorithm. In the first phase of the Adaptive Rough Fuzzy C- Means algorithm, the Rough k means algorithm is used for pre-processing of UCI repository dataset and it is normally identify the outliers and removed from the dataset. After removed outliers in the dataset, the remaining data objects are clustered by Rough K-Means, Fuzzy C-Means and Adaptive fuzzy c means methods. The experimental result of the clustering algorithms and outlier detection methods are compared and also analysed using validity indexes (Davies Bouldin index, Rand index and Adjusted Rand index), performance analysis, cluster compactness and scalability methods. Further, the experimental readings show that the Adaptive Rough Fuzzy C-Means clustering algorithm outperformed the Rough K-Means and fuzzy C-Means clustering method.