Context.AI: Context-Enhanced Financial Forecasting

User Interface
Pipeline

Inspiration

For generations, traders, business visionaries, and economists have relied on age-old financial indicators for forecasting firm performance. Yet, the digital age has ushered in a new era of possibilities, one where the world's collective sentiments play an important role in shaping a company's financial future. With the pioneering capabilities of generative AI, we have designed a paradigm-shifting model that seamlessly blends the tried-and-true accounting metrics with the ever-evolving digital landscape — redefining financial forecasting.

What it does and how we built it

Context.AI streamlines the complex task of modeling large and unstructured real-world data. The user begins by entering the name of a company in a search bar, and Context.AI begins scraping the web for the company’s financial figures and reports to leverage for prediction. Specifically, the model utilizes the Yahoo Finance and SEC APIs to search for metrics such as Revenue, EBIT, and Stock Closing Price, among others. In addition to these traditional financial metrics, the model makes use of the Google News API to scrape news articles related to the company and utilizes the FinBERT architecture to perform sentiment analysis and quantify positive, neutral, and negative sentiment about a company over time. Finally, the model also uses the Federal Reserve Economic Data API to collect macroeconomic statistics in the form of Unemployment Rate, GDP, CPI, USD Value, and CCI. When this data mining process is complete, the model utilizes vector autoregression to forecast each selected metric of company performance into the future. By drawing upon both traditional and non-traditional data sources, the model leverages contextual information about consumer sentiment and the health of the global economy.

We further design an interactive UI which allows users to precisely analyze the forecasting of each predicted company metric in order to make more informed investment decisions. Our focus on situating financial trends in a global context for our users differentiates our solution from other solutions in the problem space. Our unique and comprehensive contextualization of regional, national, and global trends allows us to exploit the synergy of structured and unstructured financial data to create more well-informed predictions.

Challenges we ran into and what we learned

This was our first time building a full-stack web application and using many of the API's and frameworks that our software makes use of. We learned about many key financial metrics and concepts as we brainstormed as well as the models and tools used to analyze financial data and unstructured text data. Learning about and implementing many of these tools and technologies within the short frame of a hackathon was a challenging but ultimately rewarding experience.

What's next for Context.AI: Context-Enhanced Financial Forecasting.

We hope to expand Context.AI's natural language processing ability to include detailed analyses of current sentiment on a particular company in its forecasting as well as the use of generative AI tools to summarize and provide guidance to users. Further, we hope to continue to develop our user experience to provide more in-depth presentation of the statistics that we collect and ultimately contribute to our final prediction.