XAI Empowered Dual Band Wi-Fi Based Indoor Localization via Ensemble Learning

Wi-Fi technology is widely used in indoor positioning systems due to its ubiquitous presence in almost every building and its cost-effectiveness without requiring additional hardware. To mitigate the effects experienced by wireless networks, dual-band Wi-Fi studies have gained importance. In this study, the UTMInDualSymFi dataset is utilized to evaluate the performance of single-band and dual-band Wi-Fi localization using 2.4 GHz and 5 GHz Wi-Fi data. For localization, KNN (K-Nearest Neighbor), XGBoost, Decision Tree, and Random Forest techniques are used for classification, and a multi-view ensemble learning approach is proposed for increasing accuracy. The results are evaluated using explainable neural network models: SHAP (SHapley Additive exPlanations) and LIME (Local Interpretable Model-Agnostic Explanations), and the effectiveness of single-band versus dual-band localization is assessed, along with the contribution of each access point to localization accuracy.