Stock volatility as an anomalous diffusion process
Year: 2024
Type: article
Abstract: # Anomalous diffusion stocks In financial markets, asset volatility—whether historical or implied—is crucial for investors. This article introduces a novel methodology for estimating stock or similar asset volatility, based on the principles of anomalous diffusion and employing Recurrent Neural Networks (RNNs). Specifically, it utilizes a combination of convolutional and bidirectional LSTM architecture within an RNN, named UPV-MAT (Universi-tat Polit ́ecnica de Valencia - Matem ́atica Aplicada), initially developed for the AnDi-challenge. This model estimates the diffusion exponent as a measure of volatility and categorizes market movements into one of four models: ATTM (Annealed Transit Time Model), CTRW (Continuous Time Random Walk), FBM (Fractional Brownian Motion), LW (Levy Walk), and SBM (Scaled Brownian Motion). We conclude that the exponent derived from anomalous diffusion processes, as determined by the UPV- MAT model, offers valuable insights into stock market volatility. It provides a novel perspective on market dynamics by elucidating whether a stock undergoes sub-diffusion, super-diffusion, or normal diffusion, thereby surpassing conventional volatility metrics.
Authors: R. V., Arevalo, J. Alberto, Conejero, Oscar, Garibo-i-Orts, A., Peris
Language: English
Field: Decision Sciences
Domain: Social Sciences
Open Access status: green
OpenAlex ID: W6968308887
FWCI:
Cites:
Cited by:
Related to: