GSM Fingerprint Untuk Deteksi Lokasi Dalam Gedung Dengan Menggunakan Algoritma Naive Bayes

Mostly, researches in indoor localization employ short wavelength signal, such as WiFi, Bluetooth, ultrasound, and infrared. In this research, indoor localization using GSM data is discussed. The use of GSM has advantages in broader coverage area and the ability for being used in power shortage condition. Indoor localization using Receive Signal Strength (RSS) GSM fingerprinting is conducted in the hallways of the third floor in EE and IT Department building, Gadjah Mada University. Fingerprinting is consisted of training phase and positioning phase. The training phase data is collected on each predefined reference points with two stepping distances (i.e., one and two meters) and constant heading direction. For the positioning phase, the data collection is performed continuously by walking along the corridor. Naïve bayes and Nearest Neighbor algorithms are used in location estimation in the positioning phase. Performance comparison of both algorithms is evaluated. The result shows that location estimation with one-meter resolution using Naïve bayes yields better result when compared with Nearest Neighbor algorithm. The minimum average error rates are 10.71 meters for Naïve Bayes(NB) and 19.64 meters for k-Nearest Neighbor (k-NN)