Skip to content

Forecasting Implied Volatility Smile Surface via Deep Learning and Attention Mechanism

Bibliographic record. Follow the original-source link for the publication.

Field Value
Primary domain Volatility
Other domains
Methods Financial Ml
Facets
Authors Shengli Chen, Zili Zhang
Published 2019-12-23
Source arXiv Quantitative Finance History
Identifiers arxiv:1912.11059
URL Open original source

Editorial synthesis

Why it matters

The study adds attention to an LSTM for implied-volatility-surface forecasts and connects those forecasts to author-reported strategy metrics, making incremental out-of-sample value the central question. (abstract:S2, abstract:S5, abstract:S6)

Main author claims

  • The authors report: An LSTM network augmented with an attention mechanism can effectively forecast implied volatility smile surfaces. (abstract:S2, abstract:S5)
  • The authors report: The forget gate of LSTM provides strong generalization and captures the long memory of financial volatility. (abstract:S3)
  • The authors report: Trading strategies constructed using the predicted volatility surfaces yield higher returns and Sharpe ratios than strategies that do not use prediction. (abstract:S5)

Data, method, or discussion scope

The evidence comes from an experimental comparison of trading strategies that use predicted versus non-predicted volatility surfaces. The approach combines deep learning (LSTM) with attention mechanisms. (abstract:S2, abstract:S4, abstract:S5)

Main limitations

The abstract does not specify the data source, market, time period, or the out-of-sample testing procedure. The risk of overfitting in the deep learning model and the incremental benefit of the attention mechanism are not quantified. (abstract:S4, abstract:S5, abstract:S6, abstract:S7)

Relationships

  • None recorded.