A Data Warehouse Solution for Flow Analysis Utilising Sequential Pattern Mining
Authors
Wejdling, Rune Leth ; Tinggaard, Simon Nicholas Moesby
Term
2. term
Education
Publication year
2009
Pages
60
Abstract
Sporing af besøgende i indendørs områder har mange anvendelser, især ved drift af store offentlige steder som lufthavne og banegårde. Denne afhandling udvikler en data warehouse-løsning (et datalager) til at lagre og effektivt analysere meget store mængder sporingsdata. Løsningen er målrettet et datasæt fra Blip Systems A/S, der sporer besøgendes bevægelser i Københavns Lufthavn. Vi anvender sekventiel mønstermining (en metode, der finder hyppige rækkefølger af hændelser) til at forhåndsberegne information om personflow gennem lufthavnen. Vi præsenterer tre måder at integrere disse hyppige bevægelsesmønstre i datalageret til flowanalyse. Et prototypesystem er implementeret som grundlag for en omfattende eksperimentel undersøgelse, der afdækker styrker og svagheder ved de forskellige tilgange og viser, i hvilke situationer de er mest anvendelige.
Tracking visitors in indoor spaces has many uses, especially for managing large public areas such as airports and train stations. This thesis develops a data warehouse—a central repository designed to store and efficiently analyze very large volumes of tracking data. The solution targets a dataset from Blip Systems A/S that tracks visitor movements at Copenhagen Airport. We use sequential pattern mining (a method that finds common sequences of events) to precompute information about how people flow through the airport. We present three ways to integrate these frequent movement patterns into the data warehouse for flow analysis. A prototype system was implemented to support an extensive experimental study, which reveals the strengths and weaknesses of the approaches and shows in which situations each is most applicable.
[This abstract was generated with the help of AI]
Documents