—Mobile location estimation is becoming an importantvalue-added service for a mobile phone operator. Itis well-known that GPS can provide an accurate locationestimation. But it is also a known fact that GPS does notperform well in urban areas like downtown New York andcities like Hong Kong. Then many mobile location estimationapproaches based on the cellular radio networks have beenproposed to compensate the problem of the lost of GPSsignals for providing location services to mobile users inmetropolitan areas, but there exists no general solution sinceeach algorithm has its own advantage depending on specificterrain and environmental factors. In this paper, we proposea selector method with LDA among different kinds of mobilelocation estimation algorithms we had proposed in previouswork to combine their merits, then provide a more accurateestimation for location services. And we build up a threelevelbinary decision tree to classify these four algorithms.These three levels are named as Stat-Geo level, CG-nonCGlevel and CT-EPM level. And the success ratios of thesethree levels are 85.22%, 88.45% and 88.89% respectively.We have tested our selector method with real data taken inHong Kong and the experiment results have shown thatour selector method outperforms other existing locationestimation algorithms among different kinds of terrains.
展开▼
