A Review on Cluster Creation for High Dimensional Discrete Data And Pattern Based Anomalous Topic Discovery

Computer Engineering

Generally, discovering of an abnormal data i.e. anomalies from discrete data leads towards the better understanding of atypical behavior of patterns and to identify the root of anomalies. Anomalies can be defined as the patterns that do not have normal behavior. It is also called as outlier detection. Anomaly detection techniques are mainly used for fraud detection in credit cards, bank fraud, network intrusion [15] etc. It can be referred as, novelties, deviation, exceptions or outlier. Such type of patterns cannot be observed to the analytical definition of an outlier, as unusual object till it has been integrated properly. A cluster analysis method is used to detect micro clusters formed by these anomalies. There are various methods existed for detecting anomalies from datasets which only detects the individual anomalies. Problem with individual anomaly detection technique that detects anomalies using the entire features typically fail to detect such anomalies. A method to detect cluster of anomalous data combine manifest atypical section of a small subset of features. This method uses a null model to for typical topic and then separate test to detect all clusters of abnormal patterns.

This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.