Comparison of MATLAB and TensorFlow Machine Learning Methods

Abstract

It can be found on the news, running our machines, and analyzing our browsing history. Machine learning is everywhere and with the recent widespread implementation of various machine learning software it is easier than ever for someone with basic programming knowledge to start building their own machine learning models. People who are new to machine learning may wonder what the various types are, what their differences are, and how they can start doing their own machine learning projects. The present study utilizes two different common machine learning techniques to build models that can predict the location of a Wi-Fi connected device through that device’s incoming router information. The first technique being used will leverage MATLAB’s vast library of prebuilt machine learning algorithms to create a model. The second type of model is a neural network, it is built using Google Brain Team’s TensorFlow 1.6. The data obtained for the models was taken throughout Tennessee Tech’s campus. This study outlines the initial data collection process used to train the models, and the processes used to create the models. The difference between neural networks and prebuilt machine learning algorithms is examined, the results of the two different methods is displayed and analyzed, and the processes used to create the machine learning models is shown as well. The models created in this study have various potential applications if used in real time on personal devices, the models could be useful for campus tours, or even adviser location for graduate students.