Characterization of personal behavior trajectory with enhanced spherical self-organizing map

The ordinary two-dimensional Self-Organizing Map (hereafter SOM) has a well-known border effect. To avoid such limitation, several spherical SOM which use lattices of the tessellated icosahedron have been proposed. However, existing data structures for such SOMs are either not space efficient or are time consuming when searching the neighborhood. We introduce a Competitive Radial Basis Function Network (CRBFN) to reduce the computation time for updating the weight on grid. Relationships of each record are maintained by their positions in the data structure rather than by immediate neighbor pointers or an adjacency list. Because of introducing CRBFN, increasing the number of neurons can be reduced. Experiments show that the spherical SOM using our data structure runs with comparable speed to the conventional 2 dimensional SOM. In addition, a heuristic method for discovering the characteristics of personality from social data is proposed in this paper. An advantage on calculating time from ordinary algorithm for Self-organizing map and example of application for socio-data is shown. Visualizing such socio-data, the fundamental method to obtain a personal behavior trajectory is proposed.