In the current era, characterized by data-driven decision-making, the practice of forecasting holds significant importance across all sectors. The ability to anticipate future events and trends, such as stock prices, consumer demand, or weather patterns, is crucial for individuals and organizations. This thesis aims to contribute to the development of both the theoretical and practical aspects of forecasting, with a specific focus on financial time series data. These data provide a complex context for evaluating and improving innovative forecasting techniques, with a particular emphasis on financial markets. However, the knowledge gained in this financial context has broader implications and can be applied in various forecasting areas. The main issue addressed in the thesis is the improvement of the accuracy of predictive models to provide actionable information for decision-making. This involves investigating and solving fundamental challenges, including the evaluation of innovative forecasting methodologies in different areas beyond financial markets.
Nell’attuale era caratterizzata da decisioni basate sui dati, la pratica della previsione riveste un’importanza significativa in tutti i settori. La capacità di anticipare eventi futuri e tendenze, come i prezzi delle azioni, la domanda dei consumatori o i modelli meteorologici, è fondamentale per individui e organizzazioni. Questa tesi si propone di contribuire allo sviluppo sia teorico che pratico delle previsioni, concentrandosi in particolare sui dati delle serie storiche finanziarie. Questi dati offrono un contesto complesso per valutare e migliorare le tecniche innovative di previsione, con un’attenzione particolare ai mercati finanziari. Tuttavia, le conoscenze acquisite in questo contesto finanziario hanno implicazioni più ampie e possono essere applicate in vari settori di previsione. Il problema principale affrontato dalla tesi è il miglioramento dell’accuratezza dei modelli di previsione al fine di fornire informazioni utili per la presa di decisioni. Questo comporta l’indagine e la risoluzione di sfide fondamentali, inclusa la valutazione di metodologie di previsione innovative in diverse aree al di fuori dei mercati finanziari.
Improving the performance of bidirectional LSTM neural networks through data preprocessing, loss function and evaluation beyond the test set: applications In causal impact analysis / Fotia, Pasquale. - (2024 Apr 23).
Improving the performance of bidirectional LSTM neural networks through data preprocessing, loss function and evaluation beyond the test set: applications In causal impact analysis
Fotia Pasquale
2024-04-23
Abstract
In the current era, characterized by data-driven decision-making, the practice of forecasting holds significant importance across all sectors. The ability to anticipate future events and trends, such as stock prices, consumer demand, or weather patterns, is crucial for individuals and organizations. This thesis aims to contribute to the development of both the theoretical and practical aspects of forecasting, with a specific focus on financial time series data. These data provide a complex context for evaluating and improving innovative forecasting techniques, with a particular emphasis on financial markets. However, the knowledge gained in this financial context has broader implications and can be applied in various forecasting areas. The main issue addressed in the thesis is the improvement of the accuracy of predictive models to provide actionable information for decision-making. This involves investigating and solving fundamental challenges, including the evaluation of innovative forecasting methodologies in different areas beyond financial markets.File | Dimensione | Formato | |
---|---|---|---|
Pasquale Fotia.pdf
accesso aperto
Tipologia:
Tesi di dottorato
Licenza:
DRM non definito
Dimensione
2.41 MB
Formato
Adobe PDF
|
2.41 MB | Adobe PDF | Visualizza/Apri |
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.