A Weight-based k-prototypes Algorithm for Anomaly Detection in Smart Grid

Anomaly detection is a typical method to find abnormal behaviors in smart grid, where the data may contain both categorical and numerical attributes with distinct significance. The k-prototypes algorithm is one of the most common algorithms for clustering mixed categorical and numerical data, however, it does not consider the significance of different attributes towards the clustering process. In this paper, we propose a weight based k-prototypes algorithm for anomaly detection in smart grid. Specifically, we first introduce an improved cost function to measure the categorical and numerical attributes uniformly and assign the weight to each attribute. We also propose two entropy metrics to calculate weight values and embed them into the k-prototypes algorithm for mixed data clustering in smart grid. Finally, we compare our proposed algorithm with existing clustering algorithms and the experimental results show that our algorithm is effective for anomaly detection in smart grid.