Random Forest Classifier in SDN Framework for User-Based Indoor Localization

Abstract

Wireless indoor localization has gained importance over the past few years. Its application ranges from turning on entertainment devices when we walk in a room to complementing biometric security systems. While present technology employing radio-frequency exists for indoor positioning, deploying these systems can be expensive and consume a lot of power. As smart systems make their way in our daily life, there is a need to develop models that can integrate with existing technology without the need of specialized hardware. We propose a random forest approach for indoor localization using a Software Defined Network (SDN) framework. This model uses random forest based cross validation to train itself and perform indoor localization. The SDN framework is responsible for coordinating the IoT devices and monitoring security threats. The system is tested using Wi-Fi signals obtained from seven different routers. The highest accuracy of 98.3% is achieved using k-fold cross validation of 3 and mtry = 2. To verify its applicability, the model was tested against other algorithms such as kNN, Support Vector Machines and Neural Networks. Random forest performs best in this scenario.