Using LDA Method to Provide Mobile Location Estimation Services within a Cellular Radio Network

Mobile location estimation is becoming an important value-added service for a mobile phone operator. It is well-known that GPS can provide an accurate location estimation. But it is also a known fact that GPS does not perform well in urban areas like downtown New York and cities like Hong Kong. Then many mobile location estimation approaches based on the cellular radio networks have been proposed to compensate the problem of the lost of GPS signals for providing location services to mobile users in metropolitan areas, but there exists no general solution since each algorithm has its own advantage depending on specific terrain and environmental factors. In this paper, we propose a selector method with LDA among different kinds of mobile location estimation algorithms we had proposed in previous work to combine their merits, then provide a more accurate estimation for location services. And we build up a threelevel binary decision tree to classify these four algorithms. These three levels are named as Stat-Geo level, CG-nonCG level and CT-EPM level. And the success ratios of these three levels are 85.22%, 88.45% and 88.89% respectively. We have tested our selector method with real data taken in Hong Kong and the experiment results have shown that our selector method outperforms other existing location estimation algorithms among different kinds of terrains.