Outlier Detection in Indoor Localization and Internet of Things (IoT) using Machine Learning

In Internet of things (IoT) millions of devices are intelligently connected for providing smart services. Especially in indoor localization environment, that is one of the most concerning topic of smart cities, internet of things and wireless sensor networks. Many technologies are being used for localization purpose in indoor environment and Wi-Fi using received signal strengths (RSSs) is one of them. Wi-Fi RSSs are sensitive to reflection, refraction, interference and channel noise that cause irregularity in signal strengths. The irregular and anomalous RSS values, used in a Wi-Fi indoor localization environment, cannot define the location of any unknown node correctly. Therefore, this research has developed an outlier detection technique named as iF_Ensemble for Wi-Fi indoor localization environment by analyzing RSSs using the combination of supervised, unsupervised and ensemble machine learning methods. In this research isolation forest (iForest) is used as an unsupervised learning method. Supervised learning method includes support vector machine (SVM), K-nearest neighbor (KNN) and random forest (RF) classifiers with stacking that is an ensemble learning method. For the evaluation purpose accuracy, precision, recall, F-score and ROC-AUC curve are used. The evaluation of used machine learning method provides high accuracy of 97.8 percent with proposed outlier detection methods and almost 2 percent improvement in the accuracy of localization process in indoor environment after eliminating outliers.