Dory | Devpost

Inspiration

Since data over time was not provided, we were unable to use a typical time series forecasting model to predict concrete values of future sales. But! We may have a solution!!

What it does

When you key in a certain company and its parameters, such as employee count and sales, you will get the projected company performance for the next 5 years. Depending on the industry, companies will be classified into their respective performance categories.

How we built it

Blood sweat and tears. Including an attempt at NLP, linear regression and identification of the R_squared value.

Challenges we ran into

Man, where do we start? Most of us lack a strong foundation in data analysis and machine learning, so even though the datathon was meant for beginners, it was really an uphill struggle. Right off the bat, we were entangled in a brutal skirmish against the 29,000+ lines of data. We have never handled such a large ginormous beast!! >.< It was so bad that our excel started lagging. Despite this excel lag and our brain lag, we brandished our swords and decided to fight this data head on together as one >:o) Since there were so many industries, we went back and forth between oversimplifying and underprocessing our data.

Accomplishments that we're proud of

We felt like giving up so many times but we are finally submitting something! We are proud that despite some of our failures, we learnt so much through this datathon and we will definitely participate in more in the near future.

What we learned

Through this treacherous journey, we have learnt new machine learning algorithms, such as k-nearest neighbours, and applied it to real-world data, which we would have never gotten the chance to. It was also through this real-world data that we realised and understood that data cleaning and wrangling is actually not as easy as we thought. Instead of just deleting data, there is many better ways to preserve the value and significance of the data, such as imputing the data.