JISE


  [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12]


Journal of Information Science and Engineering, Vol. 39 No. 5, pp. 1185-1207


SADEM: An Effective Supervised Anomaly Detection Ensemble Model for Alert Account Detection


Hui-Kuo Yang+, Bing-Li Su and Wen-Chih Peng
Department of Computer Science
National Yang Ming Chiao Tung University
Hsinchu, 300093 Taiwan
E-mail: hgyang@gmail.com
+; billy4195.su@gmail.com; wcpeng@cs.nctu.edu.tw


Anomaly detection has been an important research topic for a long time and has been applied to many real-world applications. However, due to the high cost of manually getting the instance label, researchers mostly resort to unsupervised or semi-supervised learning approaches. The supervised learning method has rarely been used in anomaly detection tasks. In this paper, we proposed a supervised learning ensemble method to detect alert accounts among transaction data. We solve the problem of low-confident predictions when the anomalies reside within normal data points. The ensemble model comprises the LightGBM and Multi-layer Perceptron (MLP) to synergize machine learning and neural network models. The proposed model preserves the result of high-confident predictions and improves the performance of low-confident predictions with the new features generated from encoding the leaf node of GBDT (Gradient Boosting Decision Tree). Our experiments on a real-world dataset show the effectiveness of the model when compared with the state-of-the-art methods.


Keywords: alert account detection, anomaly detection, imbalanced classification, supervised learning, low-confident predictions

  Retrieve PDF document (JISE_202305_11.pdf)