mundimoto

Inspiration

We were inspired because of our love for motorcycles, as well as the words shared with members of the mundimoto team during the early hours of the competition.

What it does

Our aim was twofold... to build a regression model that predicts how long a motorcycle will remain in stock without being sold (with the aim to help mundimoto plan their purchases more effectively and save resources). Our second objective was to build a recommendation system that captures the Riding feel of the bikes and recommends similar experiences (or intriguingly different experiences) that might be of interest to the user.

How we built it

We have built a thorough data scraper with the aim to make our dataset more robust. The scraper automatically extracts all motorcycle data available of the years selected by user from the bikes.com website . Then we performed some heavy preprocessing to make the new data compatible with the dataset provided by mundimoto in order to perform reconciliation. Once both datasets were reconciled, we concentrated on performing EDA and modeling the regressor for days in stock. Our efforts at building the recommendation system were halted by time, but we believe there is a lot of potential in this reconciled dataset.

Challenges we ran into

Data wrangling real data is hard (we already knew this but it never hurts to be reminded)!

Accomplishments that we're proud of

We are proud of the scraper we have built. A scraper that allowed to put together a very thorough dataset, with attributes that we feel do. great job at capturing the essence of a motorcycle and the experience it provides. Had we been able to implement the experience recommendation system, we believe this project would be out of this world. Nevertheless, we must be proud of the work produces as well as the challenges overcome.

What we learned

We learned a lot about effective time management within a group in a high pressure environment that will definitely be applied next year at HackUPC-2023. We also learned some amazing web scraping techniques to make extractions run faster and produce cleaner results.