Figure - uploaded by Apostolos G. Katsafados
Content may be subject to copyright.
Out-of-sample performance of bank stock returns prediction, using both SVD-100 textual features from press conferences of central bankers and financial variables
Source publication
We combine machine learning algorithms (ML) with textual analysis techniques to forecast bank stock returns. Our textual features are derived from press releases of the Federal Open Market Committee (FOMC). We show that ML models produce more accurate out-of-sample predictions than OLS regressions, and that textual features can be more informative...
Contexts in source publication
Context 1
... the three models achieve quite similar performance according to the results in In Table 5, we show the predictive ability of our models including both financial variables and textual features. In the empirical setting, we apply the singular value decomposition (SVD) dimensionality reduction technique (Degiannakis et al., 2018;Katsafados et al., 2023a). ...Context 2
... according to Table 5, we observe that the OLS and SVR models do not present improved performance compared with previous results. However, the interesting point here is that the RF and MLP achieve better out-of-sample performance including the combined input of textual features and financial variables than the models using a single type of input. ...Similar publications
Numerous European and national initiatives have been launched since 2014 to reduce non-performing loan (NPL) stocks on euro area bank balance sheets. NPL ratios have fallen as a result, but very gradually, mainly thanks to sales to non-bank investors. Despite stronger market activity, prices paid by NPL investors have only improved marginally and c...
