We are inspired by iPhone's app shortcut function where it shows you shortcuts to some mostly common apps based on user's app using history. We improved the idea and built an app recommendation algorithm by analyzing users' current location and previous habits.

What it does

The model predicts what apps the users may use next by analyzing their current location and previous using habits which include app categories they were using during the same time on the previous few days.

How we built it

We cleaned, processed, merged data from app_usage_events and application_versions and selected relevant features which were: app category, location, app run-time, etc.

Challenges we ran into

Challenges we ran into included handling missing data and invalid data input, discretization of continuous time, implementation of extreme learning machine.

Accomplishments that we're proud of

We are able to visualize hourly density of users as well as cdma signal strenght at a certain location.

What we learned

cdma signal strength is different for weekend and weekdays for the same location, and the strength is different during different time of the day. ## What's next for App Recommendation Based on Location & User Habits

Collect more data for each individual to refine customized app recommendation for each person.

Built With

Share this project: