Inspiration
As the problem statement was pretty clear of what is needed to be done, our goal was to actualize the product as much as possible.
What it does
It visualizes the current and forecasted sales of phones based on the data given. It also allows you to add on to the dataset and retrain the prediction model. There is inventory tracking tab which lets you set a threshold which will alert you if your projected 3 month sales will exceed the amount of phones available by a certain threshold.
How we built it
We first cleaned up the data given and modified the different tables to merge them into a single dataset. We then analyzed the data for important features and used those features to train a linear regression model. We then ported the model into Flask making a RestfulAPI. From this, the ReactJS frontend can just pull data and retrain the model by executing the flask backend.
Challenges we ran into
At the start we struggled to gain momentum as two of our four members were relatively new to programming hence we found ourselves all working on the same problem and they had no meaningful way to contribute. We combated this by splitting ourselves into two teams, one working on the prototyping and one on the data analysis and one for realizing the product. Since the only advantage of programming experience is familiarity of the syntax, it won't be of much advantage when having to solve complex problems and think.
Accomplishments that we're proud of
Despite the short span of time, we managed to realize all the functions that we wanted to achieve without having to resort to prototyping.
What we learned
A bulk of our time was spent on trying to merge the offers and holiday fields into the dataset, however after successfully merging the fields, we realized that those features actually had no correlations with the sales whatsoever and including them actually decreased our accuracy. Hence we realized that for machine learning sometimes less is more.
What's next for TEAM 34_Singtel Sales Analysis
As this was our first hackathon we did not have much expectation but we were very satisfied with the amount of progress that we have managed to complete. We look forward to joining more hackathons in the future.
Log in or sign up for Devpost to join the conversation.