engleski

Estimation of latent factors from high-dimensional financial time series based on unsupervised learning

Unsupervised learning methods have been increasingly used for detecting latent factors in high-dimensional time series, with many applications, especially in financial risk modelling. Most latent factor models assume that the factors are pervasive and affect all of the time series. However, some factors may affect only certain assets in financial markets, due to their clustering within countries, asset classes, or sector classifications. In this thesis, high-dimensional financial time series with pervasive and cluster- specific latent factors are considered. For the assumed latent factor model, an iterative method for clustering and latent factor estimation is proposed. A model selection algorithm is also developed, based on the spectral properties of asset correlation matrices and asset graphs. Based on the estimated latent factor structures, a covariance matrix estimator is also proposed, decomposing the security return covariance into the pervasive latent factor component, cluster-specific latent factor component, and a sparse idiosyncratic risk component. The covariance matrix estimates are used in a portfolio optimization scenario, focusing on risk-based portfolios. Moreover, a new portfolio optimization method based on the risk contributions of the identified latent factors and security clusters is also developed. A simulation study with known data generating processes demonstrates that the proposed latent factor estimation and clustering method outperforms other clustering methods and provides estimates with a high degree of accuracy. Moreover, the model selection procedure is also shown to provide stable and accurate estimates for the number of clusters and latent factors. In addition, risk- based portfolios using the estimated latent factor structures are tested on datasets of asset returns from global financial markets using a backtesting approach. The results demonstrate that the clustering approach and estimated latent factors yield relevant information, improve risk modelling and reduce volatility in the out-of-sample portfolio returns.

Latent factor models ; High-dimensional data analysis ; Financial risk modelling ; Portfolio optimization

nije evidentirano