-
-
Stocky - LSTM+GARCH AI model
Inspiration
We were inspired to create Stocky after observing a growing interest in investing among retail traders, students, and young professionals. Despite this rising curiosity, we found that many of the existing platforms were either too expensive, overly complex, or failed to deliver practical and personalized insights. We wanted to change that. Our goal was to develop an accessible, intelligent platform that could empower users to make smarter investment decisions using advanced technology without requiring a finance or data science background.
What it does
Stocky is a web-based application that forecasts stock prices and volatility using a powerful hybrid of LSTM (Long Short-Term Memory) and GARCH (Generalized Autoregressive Conditional Heteroskedasticity) models. Users can analyze historical price movements, identify trends, and receive real-time alerts on potential market shifts. The app also personalizes insights based on the user’s preferences or portfolio goals, helping both beginners and experienced investors make more informed, data-driven decisions.
What it does
Stocky is a web-based application that forecasts stock prices and volatility using a powerful hybrid of LSTM (Long Short-Term Memory) and GARCH (Generalized Autoregressive Conditional Heteroskedasticity) models. Users can analyze historical price movements, identify trends, and receive real-time alerts on potential market shifts. The app also personalizes insights based on the user’s preferences or portfolio goals, helping both beginners and experienced investors make more informed, data-driven decisions.
How we built it
We developed Stocky with a HTML and Python web scrapping frontend to ensure a clean and responsive user experience. The backend is powered by Python using FastAPI, where we implemented our AI models. At the core of the prediction engine lies a hybrid model that combines LSTM for capturing long-term nonlinear dependencies and GARCH for modeling market volatility. We sourced data through financial APIs and processed it using libraries such as pandas, NumPy, and statsmodels. The application is hosted on cloud infrastructure and version-controlled through GitHub, enabling continuous updates and improvements.
Challenges we ran into
One of the biggest challenges was integrating LSTM and GARCH into a single forecasting framework that could function efficiently in real time. Financial data is often noisy, incomplete, or unpredictable, which made preprocessing and model tuning especially difficult. We also faced design challenges crafting an interface that feels intuitive yet offers the analytical depth needed by more advanced users. Additionally, optimizing the model for speed without sacrificing accuracy was an ongoing balancing act, particularly given the complexity of the hybrid forecasting algorithm.
Accomplishments that we're proud of
We’re proud to have built a fully functional prototype of Stocky that not only runs sophisticated forecasting models but also presents them in a user-friendly and visually appealing way. Integrating AI models that can predict both price and volatility, while maintaining usability and speed, is no small feat. We are also proud of the fact that Stocky is affordable and inclusive, offering powerful tools at a fraction of the cost compared to traditional financial platforms.
What we learned
Through this project, we gained valuable experience in applying AI and machine learning techniques to real-world financial problems. We deepened our understanding of both LSTM and GARCH models and learned how to engineer a hybrid approach for time-series forecasting. On the software side, we sharpened our skills in full-stack web development, UI/UX design, and cloud deployment. Most importantly, we learned how to build something meaningful that solves a real user need.
What's next for Stoky
Looking ahead, we plan to expand Stocky to support multi-asset forecasting, including cryptocurrencies and commodities. We’re also developing a mobile app version to make Stocky more accessible on the go. On the technical side, we aim to improve our model by incorporating additional data such as the VIX index, economic indicators, and sentiment analysis. As we continue refining the platform, our mission remains the same: to empower users with smart, accessible, and affordable investing tools.
Log in or sign up for Devpost to join the conversation.