AdSniper

User clusters
correlation heatmap
gamer usage
Social networker
spot game users
kmeans example

Inspiration

Advertisement is tricky. Users hate non pertinent ads and companies do not want to waste money on useless advertisements. Having data about app usage for users is a valuable insight that can revolutionize targeted advertisement in-apps (and out).

What it does

We look at, study and cluster app usage profiles for mobile users. We look at frequently used app categories and when a user uses a type of app.

How we built it

We used a machine learning technique called K-means clustering. This method is supervised. A good thing about it, is that we control the number of clusters. The advertizing company can decide how many different campaigns they want and cluster their users in these groups.

Challenges we ran into

the size of data, processing millions of lines is challenging. We wish we had Microsoft or IBM as a sponsor to use a Haddop cluster or other interface to load files into HDFS and use Hive over it.