Improving the Accuracy of Fuzzy Decision Tree by Direct Back Propagation with Adaptive Learning Rate and Momentum Factor for User Localization

Most prevailing availability of wireless networks has elevated an interest in developing a smart indoor environment by utilizing the hand held devices of the users. The user localization helps in automating the activities like automating switch on/off of the room lights, air conditioning etc., which makes the environment smart. Here, we consider locating the users as a pattern classification problem and use Fuzzy decision tree (FDT) as a knowledge discovery method to locate the users based on the wireless signal strength observed by their handheld devices. To increase the FDT accuracy and to achieve faster convergence, we came up with a novel strategy named Improved Neuro Fuzzy Decision Tree with an adaptive learning rate and momentum factor to optimize the parameters of FDT. The proposed approach can be used for any classification problem. From the results obtained, we observe that our proposed algorithm achieves better convergence and accuracy.