An Indoor Positioning System (IPS) issues regression and classification challenges in form of an horizontal localisation and a floor detection. We propose to apply the XGBoost algorithm for both tasks. The algorithm uses vectors of Received Signal Strengths from Wi–Fi access points to map the obtained fingerprints into horizontal coordinates and a current floor number. The original application schema for the algorithm to create IPS was proposed. The algorithm was tested using real data from an academic building. The testing data were split into two datasets. The first data set contains signals from all observed access points. The second dataset consist of signals from the academic network infrastructure. The second dataset was created to eliminate temporary hotspots and to improve a stability of the positioning system. The tested algorithm got similar results as reference methods on the wider set of access points. On the limited set the algorithm obtained the best results.

The task of identifying native and foreign elements and rejecting foreign ones in the pattern recognition problem is discussed in this paper. Such the task is a nonstandard aspect of pattern recognition, which is rarely present in research. In this paper, ensembles of support vector machines solving two-classes and one-class problems are employed as classification tools and as basic tools for rejecting of foreign elements. Evaluation of quality of classification and rejection methods are proposed in the paper and finally some experiments are performed in order to illustrate acquainted terms and methods.

The paper presents a Wi-Fi-based indoor localisation system. It consists of two main parts, the localisation model and an Access Points (APs) detection module. The system uses a received signal strength (RSS) gathered by multiple mobile terminals to detect which AP should be included in the localisation model and whether the model needs to be updated (rebuilt). The rebuilding of the localisation model prevents the localisation system from a significant loss of accuracy. The proposed automatic detection of missing APs has a universal character and it can be applied to any Wi-Fi localisation model which was created using the fingerprinting method. The paper considers the localisation model based on the Random Forest algorithm. The system was tested on data collected inside a multi-floor academic building. The proposed implementation reduced the mean horizontal error by 5.5 m and the classification error for the floor’s prediction by 0.26 in case of a serious malfunction of a Wi-Fi infrastructure. Several simulations were performed, taking into account different occupancy scenarios as well as different numbers of missing APs. The simulations proved that the system correctly detects missing and present APs in the Wi-Fi infrastructure.