Citation

Barakat, Mohtady Ehab Hasan Aly and Chung, Gwo Chin and Pang, Wai Leong and Lee, It Ee and Chan, Kah Yoong (2025) Study of Machine Learning-Based Indoor Positioning Systems Through Wi-Fi Triangulation. In: 17th IEEE Malaysia International Conference on Communication, MICC 2025, 27 August 2025 - 28 August 2025, Melaka, Malaysia.

Text Study of Machine Learning-Based Indoor Positioning Systems Through Wi-Fi Triangulation.pdf - Published Version Restricted to Repository staff only Download (490kB)

Abstract

The development of indoor positioning systems (IPS) has been crucial recently due to their increasing usage in various applications, ranging from human or asset tracking to emergency response. By leveraging existing technologies such as Wi-Fi and Bluetooth, IPS becomes possible to be implemented easily in indoor spaces with lower setup costs. However, interference can degrade its performance drastically in densely populated areas. As such, various machine learning techniques have been introduced to overcome this challenge. The effectiveness of these models is still under research since there are various factors that affect the models’ performance. Hence, this paper focuses on developing various machine learning models for IPS through Wi-Fi triangulation. These models include the support vector machine (SVM), K-nearest neighbor (KNN), random forest (RF), neural networks, and long shortterm memory (LSTM) recurrent neural networks (RNN). Each algorithm’s performance was tested within a commercial building under several predefined configurations to evaluate their strengths and weaknesses practically. Performances were evaluated using a confusion matrix, accuracy, and F1 score. The results obtained indicate that the RF, neural networks, and LSTM-RNN models have the best performance under a fixed triangular configuration with an accuracy and an F1 score of 0.99. This strongly suggests that a fixed triangulation setup is the optimal approach for maximizing model accuracy in IPS. All these three models are also found to be more robust to changing environments, while LSTM-RNN is more exceptional.

Item Type:	Conference or Workshop Item (Paper)
Uncontrolled Keywords:	Machine learning, indoor positioning, Wi-Fi triangulation
Subjects:	Q Science > Q Science (General) > Q300-390 Cybernetics
Divisions:	Faculty of Artificial Intelligence & Engineering (FAIE)
Depositing User:	Ms Rosnani Abd Wahab
Date Deposited:	22 Dec 2025 07:22
Last Modified:	26 Dec 2025 08:54
URII:	http://shdl.mmu.edu.my/id/eprint/15121

Downloads

Edit (login required)