Improved Wi-Fi Fingerprinting Positioning Algorithm Based on Access Point and Neighboring Point Selection

Wi-Fi fingerprint-based positioning has been widely deployed for indoor localization due to its low deployment cost and good environmental adaptability. Nevertheless, positioning accuracy is significantly degraded by signal interference, redundant access points (APs), and unreliable neighboring-point (NP) selection. To address these drawbacks, this paper proposes an improved Wi-Fi fingerprint positioning algorithm integrating stepwise AP screening and adaptive NP-selection strategies. In the offline phase, preliminary AP filtering is performed according to missing-rate statistics, where the optimal threshold is determined via information entropy. An enhanced genetic algorithm (GA) combined with five-fold cross-validation and a voting scheme is adopted to optimize AP combinations. In the online phase, test points (TPs) are classified using fingerprint distance statistics, and dedicated outlier removal and weighted fingerprint distance modules are activated to select reliable NPs for different TP categories. Weighted K-nearest neighbors (WKNN) is then applied to calculate target coordinates. Experimental results demonstrate that the presented algorithm reduces the mean positioning error to 0.830 m, corresponding to a 33.81% error reduction compared with conventional WKNN. The stepwise AP selection and adaptive NP selection modules achieve error reductions of 19.70% and 25.68%, respectively, which verifies the effectiveness and robustness of the proposed scheme.