Implementation of numerical attribute discretization for outlier detection on mixed attribute dataset

This research aims to implement numerical attribute discretization for outlier detection in mixed-attribute data. We use Z-Discretization (ZDisc) technique and clustering-based discretization using K-means for the discretization and implement Attribute Value Frequency algorithm for outlier detection in the final form of the mixed-attribute dataset. We compare this technique another outlier detection method. The result shows that A VF with Z-Discretization give slightly better result compared with MixCBLOF and K-Means Discretization. It also gives an advantage that AVF with Z-Discretization has no requirement for determining initial parameter like another algorithm.