Optimum partition in flight route anomaly detection

Abstract

Anomaly detection of flight route can be analyzed with the availability of flight data set. Automatic Dependent Surveillance (ADS-B) is the data set used. The parameters used are timestamp, latitude, longitude, and speed. The purpose of the research is to determine the optimum area for anomaly detection through real time approach. The methods used are: a) clustering and cluster validity analysis; and b) False Identification Rate (FIR). The results archieved are four steps, i.e: a) Build segments based on waypoints; b) Partition area based on 3-Dimension features P<sub>1</sub> and P<sub>2</sub>; c) grouping; and d) Measurement of cluster validity. The optimum partition is generated by calculating the minimum percentage of FIR. The results achieved are: i) there are five partitions, i.e: (n/2, n/3, n/4, n/5) and ii) optimal partition of each 3D, that is: for P<sub>1</sub> was five partitions and the P<sub>2</sub> feature was four partitions

Detection Of Event-Related Potential Artifacts Of Oddball Paradigm By Unsupervised Machine Learning Algorithm

“…Supervised learning uses labeled data to help predict outcomes. On the other hand, unsupervised learning does not use labeled data [38]. They analyze and discover hidden patterns and return the data points with abnormal behavior.…”