Location prediction model using Naïve Bayes algorithm in a half-open building

Currently, received signal strength (RSS) becomes an interesting method for indoor location prediction. Some advantages using RSS compared to other methods are it does not require transmitter and receiver synchronization and it allows construction without any additional devices. There are two phases to make a location prediction, offline and online phase. Offline phase uses training data to build a fingerprint database. Then online phase is used to test the fingerprint database that has been built in the offline phase. After finishing both phases, average error between the actual distances in real coordinate points and the coordinates predicted is calculated. By performing several stages of testing, the result shows that the best prediction obtained by combining data from fingerprint database, Netsurveyor's log, and software implementation. In this way, the average error distance can be reduced up to 3 m. This is caused by lacking of variety of data retrieval with the diverse conditions.