Hybrid Modelling Based on SVM and GA for Intelligent Wi-Fi-based Indoor Localization System

Due to the growth of the ubiquitous positioning-based services, the development of indoor positioning systems has attracted the researches intensely. WiFi is one of the technologies introduced to support the indoor positioning services. The WiFi signal strengths directed from advice localized in an indoor environment can be utilized to predict the device's location the user handling that device. The prediction of the user location can be considered as a classification problem where the model can predict the location of the user according to predefined zones. Machine learning (ML) techniques have been applied widely in the literature to develop indoor positioning systems. However, these applications suffer from poor generalization ability and/or high computational complexity. This paper proposes an indoor positioning system based on a hybrid ML model that uses support vector machine as a classifier tool. To improve predictive capability of the model, SVM's parameters are optimized using genetic algorithms. The proposed model demonstrates promising results in terms of significant correlation (R2=0.99) and high classification accuracy rate (ACC= 98.3%).