Indoor Positioning Using Wi-Fi RSSI Based on LSTM-XGBoots Combined Model

In recent years, the rapid development of mobile communications and IoT technology has driven the widespread adoption of indoor location-based services, while also creating an urgent need for higher-precision positioning technology. Among various indoor positioning schemes, Wi-Fi fingerprint positioning has been widely adopted for its superior performance, but traditional RSSI-based methods still require improvement in accuracy. To address this problem, this paper proposes an indoor localization method based on a combination of LSTM and XGBoost models to improve the accuracy of RSSI fingerprint localization. The method first preprocesses the raw RSSI data using Kalman filtering to eliminate outliers and improve data reliability; then, by constructing sequential data through a sliding window, a bidirectional LSTM model is used to capture the temporal dynamic characteristics of the RSSI signals and extract deep features; finally, XGBoost is employed to perform secondary modeling on the extracted features to predict coordinates through regression, thereby achieving higher positioning accuracy. Overall, the method outperforms traditional approaches in terms of positioning accuracy and stability.