Indoor Localization Using a Hybrid Algorithm of BiLSTM and SVM

In the networking community, localization is a broadly discussed subject. It can be divided into indoor and outdoor localization, with indoor localization being particularly important for the development of smart cities. Unlike outdoor localization, which frequently uses GPS, indoor localization is difficult due to obstructions like walls, thus, a novel indoor localization model is required, which forms the main subject of our investigation. This study aims to accurately predict smartphone locations in interior situations using a deep machine learning hybrid model. To do this, we compared the accuracy of recurrent neural network methods (BiLSTM) and machine learning techniques (SVM). Our research showed that SVM demonstrated remarkable result across all devices, while BiLSTM outperformed in terms of accuracy. Based on these findings, we created the BiLSTM-SVM model, a hybrid model that combines the benefits of SVM and BiLSTM in a multi-stage collaborative framework. We used a variety of cellphones and access point models to collect the RSSI dataset for our investigation. To guarantee that the dataset is compatible with the models being used, we carried out a number of preprocessing processes. Models, such as SVM, Bi LSTM and the proposed BiLSTM-SVM model were tested for performance, then the proposed model outperformed the others. We also investigated the presence of human body inside the room. In scenarios without human presence, our suggested hybrid model, Bi LSTM-SVM, attained an accuracy of 89.11% and 78.7%, respectively. In all scenarios, the proposed model is better than those models presented in this paper.