GreenBull

Visualization of the Decision-Tree Regression ML model

Inspiration

ESG investing is a concept that should, in theory, be widely adopted by traders in the market. However, current ESG solutions for average investors are severely lacking. One of our teammates, Ahad, saw this firsthand when he interned at Lallic Partners LLC (www.lallicpartners.com) in the ESG investing division, managing 2 million USD+. Sustainable investment is extremely difficult for the average investor but is widely available for institutional investors, and the gap is only widening. Furthermore, available investment apps don’t offer ESG investing in their platform. The final nail in the coffin lies in the fact that ESG ratings are often biased, with some oil companies having high ratings due to lobbying and their claims to “become net-zero in the future.” In fact, In some ESG datasets, we found over 63% of oil companies had an A rating or above (the highest possible rating).

Well never fear, because GreenBull is here to solve all these problems! Through GreenBull, we seek to convert this convoluted, difficult and bureaucratic process into an simplistic, profitable and rewarding one for the common consumer.

What it does

GreenBull involves three aspects: Using ML to create fair ESG ratings using Yahoo Finance API data. Using a customer quiz to determine their personal ESG preferences and utilizing that to personalize stock recommendations. Using behavioral design to create a UI/UX chatbot that surveys users to find their ideal ESG portfolio

GreenBull uses Machine Learning through Decision Tree Regression to predict the ESG values of a company giving the sector, company, market cap, earnings growth, and other factors and removing bias from the process. The model accounts for these factors, meaning oil companies can’t get away with high ratings with their claims for future change.

We then survey users based on ESG preferences, developing the ideal portfolio which balances the personalized ESG ratings with returns. Lastly, our chatbot UI/UX allows users to seamlessly complete the survey and invest in the market (with a setup time of fewer than 10 minutes).

How we built it

We utilized object-oriented programming in order to create a flexible question framework that allowed us to easily add or modify questions in the future. Each question allows us to better understand the user’s preferences and find companies that subscribe to their views.

The program takes in the user's input, and uses them to narrow down to a set of 3 industries which the user prefers, as well as using their answers to calculate the amount of weight they place in each section of Environment, Social, and Governance.

Using machine learning, we fed it information using the Yahoo Finance API from a representative subset of 10,000 companies, and had it construct a model using Decision Tree Regression. Using the model, we can predict ESG values without the influence of bias, to increase ESG rating accuracy.

Finally, based on the user’s ESG preferences and the predicted ESG values, we provide the 3 best companies to invest in that align most with the user’s ESG preferences and have the greatest earnings growth among those companies.

Challenges we ran into

Firstly, a major issue we faced was preparing the data for the machine learning algorithm.

Research what factors impact ESG ratings.

Drawing the data from Yahoo Finance API took too long, due to retrieving the information off the internet. We solved this issue by training the module with a representative subset of data as a proof of concept. Given more time, we could train an even more accurate model.

There was a challenge with connecting the frontend of Figma with our complex backend. However, we mitigated this shortcoming by implementing the UI (The Questions from the customer survey) directly in the terminal.

Accomplishments that we're proud of

Successfully training a machine to predict values based on data obtained from an API was difficult and time consuming. Furthermore, there were countless bugs and errors with the internet but we persevered.

One of our members became sick during the event, but we were able to overcome this challenge by each taking on a greater number of tasks, and knowing we were able to create a successful prototype is very fulfilling.

Our skill sets are diverse, with some of us specializing in design (Figma, Frontend) and others specializing in customer analytics (ML, Programming backend). Therefore, we are most proud of the fact that we were able to take this mish-mash of skills and create something that's greater than the sum of its parts. Also, with 3 out of 4 members having never attended a hackathon before, being here, working through the night, and creating a project we are proud of is one of the best things ever.

What we learned

We learnt a lot researching ESG and social impact investing, gaining an interest in a new field of fintech.

We advanced our knowledge of ML technologies and modules. Especially regarding data preparation and utilizing APIs. Finally, we learnt a significant amount about Figma and building interactive UI.

What's next for GreenBull

We seek to improve the model by increasing training samples (20K+) Broaden the range of companies suggested by the algorithm. Fully connect the UI/UX to the backend. Integrate AI in the stock recommendation process