In the financial field, efficient portfolio management is a crucial task for investors and asset managers, since it seeks to maximize returns and minimize the risks associated with investments. In recent years, the field of artificial intelligence and machine learning has experienced remarkable progress, which has allowed the application of innovative techniques to improve the financial decision-making process.
This work focuses on the integration of two powerful tools: artificial neural networks and quadratic programming, to address the challenge of portfolio management. The combination of these techniques offers a robust and promising approach to financial time series forecasting and portfolio composition in a highly dynamic and complex financial environment.
The development of the work is structured in several fundamental sections to comprehensively address the topic. First, a detailed characterization of the financial time series is carried out, examining its essential characteristics and properties to better understand the behaviour of asset prices.
Next, the potential of artificial neural networks in time series forecasting is explored. Background on the use of these networks in this context is presented and two widely used architectures are highlighted: convolutional neural networks and Long Short-Term Memory (LSTM) networks, both with the ability to capture complex patterns in financial data.
The section on portfolio composition addresses the problem and presents various techniques applied in asset management. It is here where quadratic programming is introduced as a relevant and efficient tool for the optimal construction of investment portfolios.
Obtaining accurate and relevant data is crucial for any financial analysis and work with Machine Learning algorithms. The methodology applied to obtain data is described, and how some of the most common indicators used in finance were computed to use as descriptive variables of the problem in conjunction with historical data. It also exposes how the vectors that will be used in the modelling and training of neural networks are structured.
In the last sections, the modelling and training process is discussed, which involves the proper configuration of the neural networks and the implementation of quadratic programming to obtain optimal results. Finally, the results obtained are presented, including the predictions generated by the neural networks and the composition of recommended portfolios, thus demonstrating the effectiveness of the proposed methodology in the management of financial portfolios.
Taken together, this work seeks to provide a comprehensive and updated view of the use of artificial neural networks and quadratic programming in portfolio management, highlighting their potential as an option to improve financial decision-making and provide investors with a valuable tool for Optimize your investment strategies in a changing and competitive environment.