We wanted to see whether or not we could predict a meme's popularity before posting it.
What it does
It scrapes certain Subreddits from the popular forum Reddit for submissions made by certain demographics. We then run a machine learning algorithm on the submission text and see if any inputs made by the marketing team to our app will resonate with the chosen demographics. Any banner ads or images submitted will be run through Microsoft Cognitive Vision APIs to suggest tags. Any text on the ad will also be run through OCR and added as tags as well.
How we built it
We built the app using Express.js and SKLearn (a machine learning library for Python). It is deployed on Microsoft Azure.
Challenges we ran into
Adjusting the models to output usable data for every single demographic.
Accomplishments that we're proud of
We managed to integrate many different platforms and APIs to made a relatively polished app.
What we learned
We learned how to integrate machine learning and market analytics and that mathematical modeling is hard!
What's next for Market Gauge
Refining of the modelling of each demographic to improve accuracy.