The use of trajectory analysis on a structured SOM for behavior analysis
Date of Publication
Master of Science in Computer Science
College of Computer Studies
Judith J. Azcarraga
Defense Panel Chair
Nelson M. Marcos
Defense Panel Member
Judith J. Azcarraga
Trajectory analysis is important for discovering patterns from a time series data. It involves tracking of series of occurrences that will eventually lead to a specific event. It helps in understanding the situation of a particular domain. Self- Organizing Map (SOM) is a type of a unsupervised clustering algorithm that can be used in analyzing high-dimensional data. SOM is an artificial neural network algorithm that is most commonly used for identifying clusters, analyzing and discovering patterns in an input space and detecting correlation between these features. A Structured SOM, which is a variation of a SOM, is presented which was used to analyze and visualize trajectory patterns of a given sample. In a Structured SOM, before training, some areas of the map would be preassigned to specific class labels. During the training phase, samples are restricted to select its best-matching unit among the predened labels. This research presents the use of trajectory analysis on a Structured SOM for analyzing the patterns extracted from time-series data and using these extracted patterns to predict the behavior of new samples. For the experiments of this research EEG or brainwaves signal data was used. The research made use of three different trajectory similarity measures namely: cluster similarity, structure similarity and coordinate distance for determining the best matching trajectory. Results of the different experiments shows that using any of the thee trajectory similarity measure would yield similar results.
Archives, The Learning Commons, 12F Henry Sy Sr. Hall
1 computer optical disc ; 4 3/4 in.
Sanhi, C. O. (2016). The use of trajectory analysis on a structured SOM for behavior analysis. Retrieved from https://animorepository.dlsu.edu.ph/etd_masteral/5279