#10 - Fespresso Portfolio Allocation Model

Most classic algorithmic trading models based on machine learning can provide predictions about the market but are incapable of providing specific advice—that's exactly what our model does.

Comment 2

Inspiration

Deep Reinforcement learning algorithms are well suited for stock investments. So we try to build a web app for everyone to share the benefits of AI.

What it does

It uses the latest Reinforcement Learning Algorithm to learn the model from past stock prices data and performs very well. Therefore, it also provides personal investment suggestions according to user's situation.

How we built it

This project consists of front-end and back-end parts. Front-end is based on Flask. And back-end is based FinRL package. We connect them together to provide investment suggestions for everyone.

Challenges we ran into

Lack of time, sleep and machines. Connect Front-end Web and Back-end Algorithms.

Accomplishments that we're proud of

Our algorithms perform very well (Average 21% return annually) last year even with accounting for the recent decline. (You can find details in pictures attached)

What's next for Fespresso Portfolio Allocation Model

We try to add more features and make it customized for everyone. Currenly, we only support DOW-30 stocks.We are expanding SP-500. If we have better machines, we can provide more choices than that.

Share this project:

Updates