'Analysis & Prediction of Application Usage in Android Phones' is an Android Recommender System that aids in easy application switching by leveraging usage pattern data. Our project won the "Best Academic Project 2014" at St. Joseph Engineering College and we presented our findings at an IEEE Conference. 

Click to access the research paper.


UX Designer, Android Developer


Android App Development, Machine Learning, Database Modeling, Surveys


Android Studio, Adobe Photoshop, Netbeans


Users have a tendency to use applications on their smartphones in a sequence and often repeat the sequence of applications. These patterns of usage provide rich contextual information, but are not being used. Users are required to switch between applications by exiting out of an application, searching for the next one and then waiting while the application launches. for the application of their choice and then launch it. With the average number of apps used by the typical US smartphone owner being around 27, finding an app can be time consuming. it would be very beneficial to the users when provided with a recommendation system suggesting the most likey next applications they will be interested to use. 





To start with, we conducted secondary research to understand the different forms of visual impairments and the products currently available for the visually impaired.


We then identified potential users of the system and had hour-long interviews to understand a day in their life in depth. We designed interview protocols to gather data on how users currently navigate inside a building, what are the obstacles they face and how current systems may or may not address their needs. We then sorted the data from our interviews to understand the user needs, their fears and their current navigation methods.


Having analyzed all the information from our user interviews, we used the data to create User Personas to give us a holistic understanding of our target user group.


We used notes from our interviews to create the user journey map to cover scenarios of use and possible errors. The following were our key take-aways from user expectations:

1. Simple, minimal and understandable language for navigation instructions

2. Timely Feedback when system does not understand user input

3. Feedback on wrong turns or missed turns


We did a brainstorming exercise where we first designed solutions alone and then as a team. Evaluating our designs as a team, we weighed the pros and cons of each solution to decide that a solution should be implemented on a mobile device that one can receive audio instructions from. 

We then gathered different pieces of technology like bluetooth beacons, RFID and depth sensors and had the team play with them to evaluate what would be the right fit for our end user. Having tested each of the items, we then compared them on 4 terms namely accuracy, response rate, cost and security. Google Tango emerged a clear winner as it is a single device with capabilities of localization, mapping and tracking. Also, the depth sensors on the Tango device help detect obstructions in the way, which can help users avoid falling, which happens very often.


We then mapped how different components would interact and integrate to work as a single system. This helped the  the engineering team clearly understand different modules of the system and their interactions. 


Currently, we are interviewing users and their mobility instructors to understand what navigation instructions are helpful. At the same time, we have the software and hardware teams working on development of the application.

*This is a project in progress. Details will be updated with time.*

Click here to read about Beyster Bluepath on the University of Michigan MDP website.


UXAaron Tang, Asha Shenoy

Software: David Cao, Minh Tran, Jonathan Hamermesh

Hardware : Moran Guo, Sanika Kharkar