MCMiaomiao Chen

Miaomiao Chen

*This is an algorithmically generated author record based on publicly available information.

Miaomiao Chen has a total of 2 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Extending Quantile and Expectile Regression with Mixture Density Networks for Cryptocurrency Modelling

    by Miaomiao Chen, Nia Chen, Jennifer Chan

    Following the Market Risk Amendment to Basel I in 1996 and the Basel III reforms in 2013, Value at Risk (VaR) and Expected Shortfall (ES) have become the regulatory standards for quantifying market risk. This paper develops tail-risk modelling using a unified mixture density framework that connects quantile and expectile modelling within a single network architecture. Classical GARCH-family models are included as parametric benchmarks, and their systematic failures in the tails motivate the proposed flexible framework. Empirical results reveal that the Asymmetric Laplace (AL) distribution used for quantile models provides tail robustness and adapts well to volatile market conditions, whereas the Asymmetric Gaussian (AG) distribution produces smoother forecasts that are particularly suitable for calmer markets. Hence, an AL-AG mixture density network (MDN) is proposed with either a fixed mixing weight or a mixing weight selected on the validation set for each target level. The selected value of π balances the complementary characteristics of the two distributions at the target level. Across the empirical results, the mixture model improves extreme-quantile calibration while providing a balanced alternative to a single distributional assumption.