In this report we introduce a novel traffic model for event forecasting research that we validated against real world traffic. For the collection of reference events we used a small WSN deployed nearby various types of roads and crossroads, where the sensors transformed the sampled waveforms generated by vehicles into event descriptions that included the sensor’s location information, a time-stamp, and the intensity of the signal as a degree of confidence in its detection.
We analyzed the reference sequences of events and we proposed a probabilistic, non i.i.d. (independent and identically distributed) aperiodic traffic model, that can be used to demonstrate the effectiveness of various event forecasting protocols.
When we compared the statistical properties of the generated events to the reference, we found that the proposed model can well approximate the various descriptive statistics as well as correlation patterns of real world measurements.
Öllös Gergely, doktorjelölt, Vida Rolland, egyetemi docens, BME Távközlési és Médiainformatika Tanszék
2013. július 22.