Classification And Clustering Theory In Data Mining Based Optimization On Fingerprint Indoor Localization

Location information is precious resource. It is related to every aspect in military, medicine, industry and our daily life. Research on localization system is used to be the key technology research on national defense. Localization circuit has been implanted in all kinds of electronic products gradually. But localization went into common families when the global positioning system (GPS) and smart phones became popular. Smart user equipment introduced many needs to our life. Location based services (LBS) were born then and LBS are going to be more popular investigate points both in academy and industry. Satellite localization systems make localization service reliable outdoor. But micro wave signal cannot enter room via thick wall. The lager need, indoor localization, cannot be catered. This paper’s main topic is another localization research direction, indoor localization research. This paper investigates the signal propagation feature in WLAN (Wireless Local Area Network) and makes some summary. The details in received signal strength (RSS) based fingerprint indoor localization system is introduced. The indoor localization system is compared to a specific information retrieval system and it is divided into online phase and offline phase. The data mining theory is introduced simultaneously. Data mining theory goes well with localization system because data mining is usually used to find new knowledge in mass data set. This paper puts emphasis on the Weighed K-Nearest Neighbors (WKNN) algorithm investigation on online phase, presents the method how to find the best parameters and analyses the effect from every parameter. And it is important to reduce the computation before online phase. Fingerprint clustering on the offline phase can reduce the cost when the user queries the database. The K-means method is enhanced to fit the fingerprint data structure for indoor localization to carry out the clusters. Some evaluations for the performance of different clustering setting are given. At last an indoor localization system is implemented. It is a tiny embedded system with less hardware and all the localization calculation works with software. This system performs the similar result with theory’s. So it proves the theory’s practical value when engineering.