<![CDATA[Positioning systems have wide range of applications with the developing technology. Global Positioning System (GPS) is an efficient solution for outdoor applications but it gives poor accuracy in indoor environment. And, various methods are proposed in the literature such as geometric-based, fingerprint-based, etc. In this study, a hybrid approach that uses both clustering and classification is developed for fingerprint-based method. Information gain based feature selection method is used for selection of the most appropriate features from the WiFi fingerprint dataset in the initial step of this approach. Then, Expectation Maximization (EM) algorithm is applied for clustering purpose. Then, decision tree algorithm is used as a classification task for each cluster. Experimental results indicate that applied algorithms lead to a substantial improvement on localization accuracy. Since, cluster specific decision tree models reduce the size of the tree significantly; computational time of position phase is also reduced.]]>
Copyrights © 2016
