This project implements the paper "Extracting Movement-based Topics for Analysis of Space Use" by G. Andrienko, N. Andrienko, and D. Hecker. The project is available at https://the-breakbar.github.io/VIS2/.
The purpose of this visualization method is to analyze spatio-temporal movement patterns with the method of topic modeling. Using data which consists of points in space and time (such as cars moving through a city), the data is embedded in a voronoi diagram. This is supposed to support the exploration of spatial and temporal patterns of movement, enabling understanding of space use. However, whether this method can uncover such insights, depends on the data and the chosen parameters.
The data to be visualized has to be a .csv file in the form "id","datetime","lat","lon". Two sample datasets are provided, a dataset of car movements
in Milan, and a smaller version of the same dataset. For the computation, three configuration variables are exposed:
Max Radius: The maximum radius of the voronoi cells. A larger radius results in larger cells, and vice versa. Increasing the radius will decrease the computation time.
Number of Topics: The number of topics for the topic modelling (aka the number of colors). A higher number will result in more topics (color), but reduces visual clarity. The optimal number highly depends on the data.
NMF iterations: The maximum number of iterations for the NMF algorithm. NMF is a random algorithm, so increasing the iterations will make the result more stable. Increasing the number of iterations will increase the computation time.
The source code is available at https://github.com/The-breakbar/VIS2. The project itself is a static web page built with d3.js, webpack and typescript. All computations are done on the client side.