Real-Time Lecturer Localization via Edge-Standardized Wi-Fi Fingerprinting on Android

Purpose: This research aims to develop a highly responsive Indoor Positioning System (IPS) to locate lecturers in real-time within educational facilities by mitigating Access Point (AP) Churn through decentralized data preprocessing. Methods: A Wi-Fi fingerprinting dataset comprising 1,428 observations from 40 Access Points (APs) was collected across lecturer operational areas. Missing RSSI values were standardized to −100 dBm on the Android client before transmission. Random Forest, Decision Tree, Support Vector Machine (SVM), and K-Nearest Neighbor (K-NN) were comparatively evaluated using stratified cross-validation and validated through real-time Android deployment. Result: The Random Forest algorithm achieved the best performance with 97.90% accuracy and 97.86% F1-score. The proposed Android-based workflow enabled reliable real-time lecturer localization while maintaining consistent RSSI preprocessing between the mobile application and the cloud inference model. Novelty: The contribution of this study is an Android-based edge preprocessing framework that standardizes missing RSSI values before cloud inference and demonstrates its practicality through comparative machine learning evaluation in a real-time lecturer localization system.