Hybrid classifier for Wi-Fi fingerprinting system

Many algorithms proposed for indoor positioning using Wi-Fi based networks, those that rely on hybrid-based approach have been demonstrated to outperform those based on deterministic, probabilistic, and pattern recognition in terms of accuracy. We construct a hybrid algorithm architecture on three representative Wi-Fi fingerprinting techniques, based on the K-Nearest Neighbor (k-NN), Naive Bayes Classifier (NBC) and Support Vector Machines (SVM) algorithms. Our experiments show that hybrid-based approach outperformed all three K-NN, NBC, and SVM-based fingerprinting, achieving accuracies of 89.47% within 2 meters or better within our testbed.