End-to-end portfolio optimization: integrating LLMs and deep learning in a web application for smart and personalized investing

Bosco, Leonardo (A.A. 2023/2024) End-to-end portfolio optimization: integrating LLMs and deep learning in a web application for smart and personalized investing. Tesi di Laurea in Machine learning, Luiss Guido Carli, relatore Giuseppe Francesco Italiano, pp. 57. [Master's Degree Thesis]

[img]
Preview
PDF (Full text)
Download (1MB) | Preview

Abstract/Index

Background and motivation. Role of artificial intelligence in modern portfolio management. The need for user-centric design. Proposed framework. Literature review. Traditional portfolio theory. Machine learning in finance. Deep learning for time series forecasting. End-to-end portfolio optimization. Large language models in financial applications. Eda & data preparation. Data collection and overview. Computing daily returns and initial exploration. Constructing rolling windows. Scaling and data preparation workflow. Concluding remarks on EDA and data preparation. Methodological framework and model exploration. Conceptual overview of the end-to-end architecture. Neural forecaster: LSTM-based architecture. Allocation layer. Utility functions and performance metrics. Negative Sharpe ratio and end-to-end training. Concluding remarks on the allocation layers. Evaluation and experimental results. Portfolio evaluation function. Performance metrics computation. Comparison of our model performance with a simple Markowitz. Comparison of our model performance with a two steps model. Web-application development. Rationale for a simple, intuitive interface. Technology stack: Streamlit. Conversational interaction. Concluding remarks on web-application development. LLM integration for input and output. Introduction to Large language models (LLMs). Gemini. The application of LLM on our web app.

References

Bibliografia: pp. 55-56.

Thesis Type: Master's Degree Thesis
Institution: Luiss Guido Carli
Degree Program: Master's Degree Programs > Master's Degree Program in Data Science e Management (LM-91)
Chair: Machine learning
Thesis Supervisor: Italiano, Giuseppe Francesco
Thesis Co-Supervisor: Simeone, Antonio
Academic Year: 2023/2024
Session: Extraordinary
Deposited by: Alessandro Perfetti
Date Deposited: 02 Jul 2025 10:08
Last Modified: 02 Jul 2025 10:08
URI: https://tesi.luiss.it/id/eprint/42702

Downloads

Downloads per month over past year

Repository Staff Only

View Item View Item