The UJIIndoorLoc database is a multi-building and multi-floor database for WLAN fingerprint-based indoor localization problems. It consists of 520 different Wireless Access Points (WAPs), along with latitude and longitude, floor number, building ID, space ID, relative position, user ID, phone ID, and a timestamp. The wifi fingerprints can be characterized by the Recieved Signal Strength Intensity (RSSI), which ranges from -104dBm (very low signal) to 0dbM. A value of 100 indicates no signal was detected. The dataset consists of both a training and testing set.
To prepare data for analysis, variables with zero variance were removed. Other irrelevant features were removed as well, including latitude, longitude, user ID, phone ID, and timestamp. Data was then separated according to building number (0, 1, and 2). A unique identifier combining Building ID, Floor ID, and Space ID was created called "Location" and was used as the dependent variable. Different models were tested for each building number, with RandomForest achieving the best results.