Regression Analysis for Estimated Distance in Fingerprinting-Based WLAN Outdoor Localization System

Wireless local area network (WLAN) localization techniques are evolving in line with technological developments and the number of wireless device users. The existing localization techniques have several methods, with varying degrees of accuracy, and are generally applied to indoors. The targets sought in existing localization techniques find positions of user's mobile device. In this paper describes the regression analysis method for fingerprinting-based WLAN outdoor localization system. The position being searched is the location of the access point, rather than the user's mobile device like any other localization research. With a signal fingerprinting system, the empirical data obtained from field measurements are stored in the database. The database consists of a DataPoint table, which includes received signal strength by the finder (RSSfnd) and the distance between the finder against the access point (Dreal). Measurements were made at a range of 0 to 100 meters and divided into eleven measurement points. Regression models used for analysis are linear regression, exponential regression, and polynomial regression. Based on the regression line and the value of $\mathbf{R}^{2}$ can conclude the most precise regression technique to estimate the distance between the finder against the target of an access point. Linear regression yields $\mathbf{R}^{2}$ value of 0.8133, exponential regression of 0.8641, and polynomial regression of $\mathbf{R}^{2}$ value of 0.9951. Based on the amount of $\mathbf{R}^{2}$ obtained, the polynomial regression is the most precise regression model compared to other regression models. The system in this paper offers a more effective and efficient method of WLAN outdoor localization, only one measurement of received signal strength (RSS) has been able to estimate the distance between the finder against the target of an access point. The system in this paper does not require an anchor or reference node when estimating distances as needed in other research.