Journal of Forecasting最新文献

Forecasting corporate bankruptcy and financial distress is a crucial and intriguing research topic within the realm of business and finance. Recently, numerous models for forecasting bankruptcy and financial distress have been developed using artificial intelligence techniques under a supervised learning paradigm. However, in practical applications, generating large amounts of labeled samples for training supervised learning models is a highly inefficient and labor-intensive process. In this paper, by taking practical application scenarios into consideration, we propose a novel bankruptcy and financial distress prediction method called Class-Rebalanced Self-Training Semi-Constrained Matrix Factorization (CRST-SemiCMF), which integrates the class-rebalancing technique and the self-training strategy within a constrained matrix factorization framework. The CRST-SemiCMF method can alternatingly train a semi-constrained matrix factorization model on both labeled and unlabeled datasets, progressively expanding the labeled dataset by sampling the unlabeled dataset with pseudo-labels. Crucially, our method can adaptively determine the sampling rates for different classes of imbalanced datasets. Moreover, we also introduce a novel modification to the self-training strategy employed in our proposed method to effectively address class-imbalance issues commonly encountered in datasets for bankruptcy and financial distress prediction. To validate the proposed method, we conduct extensive experiments using multiple UCI benchmark datasets and a real-world financial dataset of Chinese listed companies. The results demonstrate that (1) our method achieves the highest classification accuracy across nearly all proportions of available labeled samples; (2) it obtains a false negative rate on par with supervised-CMF while surpassing SemiCMF; and (3) it gets better Recall and F1-scores than SemiCMF, with performance comparable to or exceeding supervised CMF. These findings consistently confirm our method's effectiveness for bankruptcy prediction and financial distress forecasting across multiple evaluation metrics.

The study estimates the extent of equity home bias (EHB) using a robust mean-CVaR optimization technique and International Capital Asset Pricing Model (I-CAPM) for the G7 and BRICS group of countries to investigate whether the latter approach overestimates the EHB. We also examine whether EHB changes over time. Our findings suggest that I-CAPM overestimates EHB for developed countries. The study also reveals a downward trend in EHB for developed countries, while it finds no significant trend for emerging countries. By using a sophisticated technique to estimate EHB more accurately, the study offers a deeper understanding of the EHB puzzle.

In this paper, the loss given default (LGD) distribution of defaulted debt is explored through the application of a cumulative link model that takes into account obligor-specific frailty. When constructing a model, it is important to recognize and address any correlations between LGD variables for defaulted debts from the same obligor. Our research highlights the significance of incorporating obligor-specific frailty into our proposed model, as it effectively captures the main characteristic of LGD variables. We demonstrate the use of our proposed frailty model through a real data example. Our empirical results support the significance of the obligor-specific frailty variable included in the proposed model. We further find that, in contrast to the independence alternatives, our proposed model achieves better out-of-time performance using an expanding rolling window approach, thereby enhancing the precision of LGD distribution predictions. The exceptional predictive accuracy of this model provides valuable insights for creditors and policymakers in assessing and managing credit risk.

Stock price forecasting has always been a classic and challenging task, attracting widespread attention from stakeholders such as market regulators, financial practitioners, and individual investors. Developing new models to improve the accuracy of stock price forecasting is also a persistent goal pursued by researchers. In recent years, quantum computing has developed rapidly. The emergence of quantum machine learning (QML) and quantum neural network (QNN) models has made it possible to develop new stock price forecasting models that leverage the advantages of quantum algorithms. To improve predictive accuracy, this study proposes a novel decomposition-ensemble approach based on multivariate empirical mode decomposition and QNN. In the prediction, multi-source big data from the stock market, search engines, and social media are employed to represent investor anticipation, attention, and sentiments, respectively. Using the daily average stock price in the Shenzhen Stock Exchange, an empirical analysis is conducted to illustrate the proposed approach. The results suggest that the proposed approach outperforms benchmark models, indicating that it is a promising method for forecasting stock price series with high volatility and nonlinearity.

Credit rating serves as a crucial instrument for lenders to evaluate borrowers' creditworthiness and mitigate the risk of nonperforming loans. However, credit rating tasks often face significant challenges due to multiclass distributions and severe class imbalances. Given the advantages of ensemble learning methods in addressing these challenges, this study presents a novel multiclass imbalance classification framework that integrates the Error Correcting Output Codes (ECOC) decomposition approach with diverse dichotomizer imbalance algorithms to enhance credit ratings. Nevertheless, selecting and quantifying the uncertainty of dichotomizer sets poses challenges. To this end, we introduce a dynamic ensemble selection strategy and evidence theory within the ECOC setup. By tailoring specific dichotomizers to individual samples and consolidating uncertain binary outcomes using belief functions, a resilient ensemble classifier is developed. Extensive experiments on nine KEEL benchmark datasets and two real credit datasets demonstrate its effectiveness in handling severe imbalance in credit rating tasks.

A positive relationship between Earning Call Sentiment and Stock Market Reaction has already been identified. Still, this utilization for prediction has yet to gain much attention. This study explores the predictive potential of earnings calls by employing the BERT model to extract key topics and their sentiments. Various machine learning techniques are then employed to leverage these insights for predicting stock market reactions and associated risks, evaluating the extent to which earnings call topics and sentiments can enhance prediction accuracy. Analyzing all quarterly earnings calls from S&P 500 companies in 2022, the results indicate that the decomposition into key topics with their respective sentiment outperforms the usage of overall sentiment across multiple scenarios and models. The random forest model is found to make the best utilization of the decomposition.