Have you ever wondered how many likes your next Instagram picture will get? What captions your followers like the best and what hashtags are trending? Well now you can, with InstaAnalytics.

What it does

InstaAnalytics takes data from your old posts and uses machine learning to find specific patterns within these older posts. Theses patterns are then analyzed and will return how many likes your next post will get.

How we built it

We built the app using a python foundation. We put up a quick flask backend server and used the clarify api to analyze the picture. A user will upload a picture and the picture will get run through our machine learning algorithm. We used the scikit-learn library to run linear regression algorithms. We gathered metadata like followers, following, and likes on previous pictures.

Challenges we ran into

Due to timing of the hackathon we could not use our initial plans of using convolutional neural networks and random forest regression algorithms. It would have taken months to feed meta data into these algorithms. Due to this we had to settle with a linear regression algorithm. We also had very limited knowledge of the Clarifai api and machine learning as a whole; so a majority of our time went to learning foundations of machine learning.

Accomplishments that we're proud of

With the lack of overall machine learning experience we were still able to put together a working application. Our application is able to take in a picture, run the machine learning algorithm through the picture and able to generate a number for the likes the picture will get.

What we learned

We learned a lot about machine learning and the algorithms that go behind them. We learned how to use flask and the Clarifai api. We learned to work together in a team in a really small amount of time.

What's next for InstaAnalytics

We are planning on really scaling InstaAnalytics into a bigger application. We are going to rewrite our machine learning algorithm to use a more complex one like convolutional neural networks and random forest regression algorithm. Gather even more meta data to get better models. Eventually figure out which hashtags and comments will generate the most likes for a user. This will eventually be moved to other different social media like Twitter and Facebook to analyze different posts.

Share this project: