danielegiulianini / inbestment-portfolio

A web-based and machine-learning fostered prototype tool to find your best financial investment portfolio

inbestment-portfolio

Table of Contents

1. About The Project
2. Documentation links
3. Solution strategy
4. Technologies
5. License

About The Project

inBestmentPortfolio is a FinTech project ¹ aimed at defining the best investment portfolio by combining to a variable extent a given set of finacial indices, over a predefined time horizon.

(back to top)

Documentation links

• Presentation slides²
• Project requirements text²

(back to top)

Solution strategy

The high-level solution strategy adopted consists in optimizing a trade-off between the estimated stock returns and the risk associated with a combination of the stock indices. This is accomplished by exploiting, among the others:

• ARIMA (GEP Box et al., 1976) model, with some experiments ARIMA in rolling window mode along with some decision tree and bagging -based regressors, leveraging grid-search for hyperparameter’s tuning, for returns estimation;
• both the Volatility Weighted Historical Simulation (John Hull et al., 1998) and the Filtered Historical Simulation (Barone-Adesi et al., 1998) approaches, for risk estimation;
• GARCH (1986) model, for conditional variance estimation, to support risk estimation;
• the Differential Algorithm (Rainer Storn et al., 1997) metaheuristic and the PSO (James Kennedy et al., 1995) genetic algorithm, for returns-risk trade-off optimization.

Python code here.

(back to top)

Technologies

The project is implemented with:

• Python
• .NET core
• Entity Framework
• sklearn
• pmd-arima
• arch
• sqlite

(back to top)

License

Distributed under the GPL License. See LICENSE for more information.

(back to top)

Footnotes

1. dating back to 2019 (during my masters's years) ↩

2. in Italian, as part of the project has been developed in the context of a university course tought in Italian that defined its requirements. ↩ ↩²