Indoor Localization Using Improved Multinomial Naïve Bayes Technique

With the extensive use of mobiles, tablets, laptops and other Wi-Fi carrying handheld devices, indoor localization using Wi-Fi fingerprinting has gained much interest of researchers. Many techniques have been introduced to increase the accuracy of the localization system. Bayesian learning techniques are considered much accurate for localization but still there are some issues including zero probability and good accuracy. In this paper we introduce a unique weighting technique called improved multinomial Naive Bayes technique for localization. For data collection we used a freeware android software, Wi-Fi Analyser. Experiments are conducted in the first floor of my office using HTC One. Our technique which uses the concept of Multinomial Naive Bayes classifier which is actually not used before in indoor localization. It provides better accuracy, resolves zero probability issue caused due to data incompleteness. It also somehow tackles with naive Bayes issue of independencies that according to Navies Bayes all the features are independent of each other but in physical circumstances it is not the case as features are dependent sometimes so we have tried to solve this issue as well and is easy to implement as it involves less computations as compared to those weighting techniques that includes non-linear functions.