MACHINE LEARNING UNTUK LOCALIZATION DALAM GEDUNG BERBASIS RSS FINGERPRINT MENGGUNAKAN IEEE 802.11g

• Main Authors: , CHAIRANI, , Widyawan, S.T., M.Sc., Ph.D.
• Format: Theses and Dissertations NonPeerReviewed
• Published: [Yogyakarta] : Universitas Gadjah Mada 2012
• Subjects: ETD
• Online Access: https://repository.ugm.ac.id/100571/; http://etd.ugm.ac.id/index.php?mod=penelitian_detail&sub=PenelitianDetail&act=view&typ=html&buku_id=57100
• Institution: Universitas Gadjah Mada

cross validation This research discusses indoor localization using the IEEE 802.11g wireless network with machine learning approach. Measurement of RSS method is based on RSS-fingerprinting. Machine learning algorithms that are used for estimating location of the RSS are kNN and Naive Bayes. In this research, indoor localization is conducted in the entire room on the 3rd floor of Electrical Engineering and Information Technology (JTETI) Universitas Gadjah Mada (UGM) building. Area of the building is 1969,68 m 2 . Four different configurations are provided in the experiments. The best configuration would be used as indoor localization in all room of the 3rd floor. The first experiment is measurement of RSS fingerprint with 2 m 2 area of grid in 3rd floor corridor, the second experiment is measurement of RSS fingerprint with 1 m 2 area of grid (the first and second experiments use all of the permanent access point (AP) on the JTETI UGM building), the third experiment is measurement of RSS fingerprint with 1 m 2 area of grid. The third experiment only uses permanent AP on the third floor. The fourth experiment is measurement of RSS fingerprint with 1 m 2 area of grid. The experiment is used two AP on the 3rd floor and three additional AP, which has set randomly on the 3rd floor. Result of the experiments show that the best accuracy is obtained from the fourth experiment, which uses RSS fingerprint with two AP on the 3rd floor and three additional AP in 1 m 2 area dimension of grid. The accuracies that are obtained using in the learning phase are 89,46% using kNN with k = 1 and 43,74% using the Naive Bayes algorithm. Average errors of distance estimation are 4,13 meters using kNN algorithm and 6,20 meters using Naive Bayes algorithm in the online and post learning phase. From the experiments, average errors of distance estimation for indoor localization in all room of 3rd floor are 4,93 meters using kNN with k=1 and 6,29 meters using the Naive Bayes algorithm in the offline and learning phase. In the online and post learning phase, average errors of distance estimation average are obtained 5,25 meters using kNN algorithm and 7,82 meters using Naive Bayes algorithm.