Human Traffic Prediction based on Limited Measurment Data

OData support
Supervisor:
Dr. Kővári Bence András
Department of Automation and Applied Informatics

The increasing competition for market share in commerce naturally raises the demand to operate businesses and shops as cost efficient as possible. The observation and measure of customer traffic data thus become an important factor in management and division of labour.

Several widespread approach exist to accomplish these goals; one of the most well-known is using photoresistor detectors. The subject of my thesis is processing movement data using an infrared motion-detector. With the measured data, it becomes possible to predict future information using prediction algorithms, which each give slightly different accuracy and efficiency.

In this thesis, I present a number of these algorithms. As the algorithms behave differently, there is a demand to measure the goodness of each algorithm. To make a proper comparison, I use different measures to numerically calculate the value of goodness. I use these results to give an analysis of each algorithm.

I made several measurements using our test environment which is thoroughly reviewed in my work. These readings showed that which are the best algorithms in different scenarios and environments and also indicated the wide number of approaches available to measure their goodness.

As an example, using the classic average proved to be one of the most efficient method in a general case. By learning the characteristics of these algorithms, we can give accurate predictions of traffic data.

Downloads

Please sign in to download the files of this thesis.