Detecting Meaningful Places and Predicting Locations Using Varied K-Means and Hidden Markov Model

This work describes the use of Varied-K Means clustering and Hidden Markov Model techniques to predict a user’s future movement based on the user’s past historical data. Several techniques [1] [2] were proposed to predict a user’s movement, but not many have concentrated on the user’s location based on both weekday and time period within the day. We have introduced a method which models the user’s data, not by just taking day of the week into consideration but also time interval. Our model is able to answer day-specific queries like "Where is the user most likely to be when it is a Monday?" or day and time-specific queries like "Where is the user most likely to be between 6:00 pm and 9:00 pm on Saturdays?" Our work gives us much higher prediction accuracy than previous research on this topic [9]. Such a model has multiple applications, which are described in the Introduction and Motivation section of the paper.