Datathon centered around finding/recommending patterns in mobile/tablet market, and investing for various app markets domain-upliftment. Refer to docs/ for more on problem statement, slides & screenshots of the approach and such. Some samples are included below:
Raw statistics, subject to answering qualms about bidding, and such
Visualizations depicting correlations and to help with extracting meaningful features meant for Phase 3
Predictions, with ExchangeBid as input, Outcome as output and the features obtained from Phase 2 as weights.
So, we started with an ipython notebook then moved on to IBM watson analytics, and finally ended up using Dato's GraphLab for logistic regression, elasticsearch & Kibana for visualizations and some spreadsheet magic to infer "Where to put my money?"
- Team Vicarious
- Comaprision of Average ExchangeBid with Age
- Comparision of Aberage ExchangeBid over Country and Outcome
- ExchangeBid% v/s DeviceType
- ExchangeBid avg. compared over Manufacturer and Outcome
- ExchangeBid avg. over countries
- Successful CampaignID(s) v/s avg ExchangeBid
- Elasticsearch indexed dataset