BROWSE
Detail
Journal Article
WiFi 신호 강도 기반의 실내 측위를 위한 머신러닝 회귀 알고리즘의 비교 분석
- Authors
- 김용현
- Issue Date
- 2024-10
- Citation
- 한국측량학회지, v.42, no.5, pp.483-490
- ISSN
- 1598-4850
- Publisher
- 한국측량학회
- Language
- Korean
- Type
- Journal Article
- Abstract
- Wi-Fi strength-based indoor positioning remains widely used across various fields because it does not require additional signal transmitters or receivers. A pre-constructed strength database of radio signals is necessary for Wi-Fi signal strength-based positioning, typically structured as a tabular dataset. Despite the rapid advancements in artificial intelligence, applying deep learning to such tabular data still results in unstable learning, with traditional machine learning-based regression algorithms outperforming in terms of positioning accuracy. In this paper, we conduct an in-depth analysis of these trends by applying various machine learning regression algorithms to a carefully constructed signal strength distribution dataset, comparing their strengths and weaknesses. The algorithms used include Extra Trees Regression, Random Forest, Support Vector Regression, and Extreme Gradient Boosting. Each algorithm’s accuracy and training time were analyzed. The experimental results show that Extra Trees Regression achieved the highest positioning accuracy, while XGBoost demonstrated the fastest training time. The dataset used in this study integrates data from all floors of a single building, organized into x-, y-, and z-(floor) coordinates. Most machine learning-based regression algorithms have demonstrated the ability to predict the x- and y-coordinates with a root mean square error of less than 2 meters, while achieving over 99% accuracy in predicting the z-coordinate(floor level).
- KSP Keywords
- Gradient Boosting, IEEE 802.11(Wi-Fi), In-depth analysis, Learning-based, Machine learning regression, Positioning accuracy, Radio Signal, Regression Algorithms, Root-Mean-Square(RMS), Signal Strength, Strength distribution
This work is distributed under the term of Creative Commons License (CCL)
(CC BY NC)
(CC BY NC)
