Abstract
With the rise in the amount of data being produced on an average day, the ability to reduce the amount of data, while still retaining its value, is crucial to analyze and visualize data quickly and accurately. In one scenario, a user decides to analyze the traffic patterns in a given space. The user sets up wireless sensors that ping the location of pedestrians at random intervals. The sensors then return a list of GPS coordinates to the user. With a large volume of pedestrian traffic, a vast given area, or an extended timeframe to gather the data, the returned information can add up to a significant amount of coordinates. In an effort to increase the speed and accuracy of analysis or visualization, the list of coordinates can be summarized. In this thesis, we propose a novel tool with a primary function of summarizing the gathered GPS coordinates, gathered by the user. Using these coordinates as input, a variant of the Kohenen Self-Organizing Maps summarizes the input, returning a reduced number of GPS coordinates back. Through testing, it is shown that various parameters, in conjunction with fractal initialization of neurons, can yield successfully summarized collections of coordinates, with results comparable to, or better than, known methods such as random sampling and K-Means clustering.