We found a match
Your institution may have access to this item. Find your institution then sign in to continue.
- Title
Hyperparameter Optimization for Indoor Localization in Wi-Fi IoT Application.
- Authors
Mane, Sarika; Kulkarni, Makarand; Gupta, Sudha
- Abstract
Wireless Fidelity (Wi-Fi) based localization suffers from multipath propagation, signal interference, and signal loss. It affects the precise distance estimation in localization. In localization based applications, accurate distance measurement is the major challenge. This research work proposes a Grid Search approach for enhancing the accuracy of indoor localization in Wi-Fi based Internet of Things (IoT) environments. Grid Search systematically explores hyperparameter combinations, to find the optimal configuration settings for machine learning models. Hyperparameter optimization is essential as it enhances model performance by fine-tuning. In this research work, three distinct models are employed: Support Vector Regressor (SVR), k-Nearest Neighbors (KNN) and Random Forest (RF). The Grid Search approach for hyperparameter tuning is employed to train each model. It utilizes Euclidian Distance (ED) measurement and Received Signal Strength Indicator (RSSI) data. The model's performance is evaluated using performance metrics. The Grid Search approach with the RF achieved an optimal Mean Absolute Error (MAE) of 0.8258 m. The validation of performance through comparative result analysis with existing research work underscored the effectiveness of the approach. In Scenario 1, remarkable optimal improvement of 54.92% is observed. Similarly, in Scenario 2 and Scenario 3, significant enhancements of 29.38% and 24.09% are obtained respectively. Grid Search with SVR showed superior performance, producing Mean Squared Error (MSE) 1.0826 meter2. This highlighted the robustness and superiority of the proposed approach in improving models performance over existing research.
- Subjects
MACHINE learning; INTERNET of things; K-nearest neighbor classification; INTERNET searching; RANDOM forest algorithms
- Publication
Wireless Personal Communications, 2024, Vol 139, Issue 4, p2601
- ISSN
0929-6212
- Publication type
Academic Journal
- DOI
10.1007/s11277-024-11727-7