basically, we saw how embarrassing our recent purchase history was with our new transferwise mastercard but our consumerist desires are still getting the best of us and we want to save up for a holiday. Fortunately we've been told that transferwise is letting us fiddle with an account worth millions.
We started dreaming with what we could spend with that, so we consulted our LSTM and it told us we love holidays. It was right, but it also told us how expensive it was...
What it does
It's quite simple:
- we built our friend to look at our historical transactions abroad.
- we then built a relatively alright formula to predict how much money i'd be spending in whatever country i guess and for how long.
- we considered currency buying power datasets and cost of living into consideration, alongside purchase histories
- then we trained a cutting edge machine learning model to predict the cost of currency pairs so we can choose the best time to convert it in time for our holiday.
- then we built a front end that takes as input some country, and some date range.
- it'll then tell you when we think it's the best time to move your currencies around (we'll even do it for you)
- and we'll even tell you how much we think you should save based on our models.
How we built it
- a rigorous and head scratching approach into building LSTMs, but we did mess around with other models such as a linear regression and decision trees
- a react front end
- flask backend
- lots and lots of country related datasets
Challenges we ran into
-Integrating UI & back-end services
Accomplishments that we're proud of
- A beautiful and simple UI
- Incoperating Transfer Wise
What we learned
- Various new technology skills and knowledge
What's next for TWT
- Budget Management