Deep K-Means Algorithm Based Autoencoder for Outlier Detection

This paper presents a new method that combines deep k-means clustering with granule mining approaches to utilise contextual information for improving outlier detection and classification. The technique of deep k-means is employed to describe features and cluster data, while granule mining is used to enhance the granularity of clusters to increase their purity. The process of identifying outliers is guided by the incorporation of contextual information. Our research enhances the field of outlier detection by highlighting the significance of contextual information and presenting a method that combines deep learning, granule mining, and contextual analysis. The experimental results on several datasets provide evidence that our suggested strategy successfully detects outliers and enhances classification performance in comparison to baseline methods. By utilising contextual information, we can get more precision in identifying anomalies and improve the overall quality of the dataset for subsequent classification tasks.