A Meta‐learning Framework for Bankruptcy Prediction

The implication of corporate bankruptcy prediction is important to financial institutions when making lending decisions. In related studies, many bankruptcy prediction models have been developed based on some machine‐learning techniques. This paper presents a meta‐learning framework, which is composed of two‐level classifiers for bankruptcy prediction. The first‐level multiple classifiers perform the data reduction task by filtering out unrepresentative training data. Then, the outputs of the first‐level classifiers are utilized to create the second‐level single (meta) classifier. The experiments are based on five related datasets and the results show that the proposed meta‐learning framework provides higher prediction accuracy rates and lower type I/II errors when compared with the stacked generalization classifier and other three widely developed baselines, such as neural networks, decision trees, and logistic regression. Copyright © 2011 John Wiley & Sons, Ltd.     

A new insight into combining forecasts for elections: The role of social media

This study is devoted to gain insight into a timely, accurate, and relevant combining forecast by considering social media (Facebook), opinion polls, and prediction markets. We transformed each type of raw data into the possibility of victory as a forecasting model. Besides the four single forecasts, namely Facebook fans, Facebook “people talking about this” (PTAT) statistics, opinion polls, and prediction markets, we generated three combined forecasts by associating various combinations of the four components. Then, we examined the predictive performance of each forecast on vote shares and the elected/non‐elected outcome across the election period. Our findings, based on the evidence of Taiwan's 2018 county and city elections, showed that incorporating the Facebook PTAT statistic with polls and prediction markets generates the most powerful forecast. Moreover, we recognized the matter of the time horizons where the best proposed model has better accuracy gains in prediction—in the “late of election,” but not in “approaching election”. The patterns of the trend of accuracy across time for each forecasting model also differ from one another. We also highlighted the complementarity of various types of data in the paper because each forecast makes important contributions to forecasting elections.     

Breast metastasis from small cell lung carcinoma

Breast metastases from extramammary neoplasms are very rare. We presented a 66 year-old female with metastasis of small cell lung carcinoma to the breast. She presented with consolidation over the left upper lobe of her lung undetermined after endobronchial or video-assisted thoracoscopic surgery （VATS） biopsy, and this was treated effectively after antibiotic therapy at initial stage. The left breast lumps were noted 4 months later, and she underwent a modified radical mastectomy under the impression of primary breast carcinoma. However, the subsequent chest imaging revealed re-growing mass over the left mediastinum and hilum, and cells with the same morphological and staining features were found from specimens oftransbronchial brushing and biopsy. An accurate diagnosis to distinguish a primary breast carcinoma from metastatic one is very important because the therapeutic planning and the outcome between them are different.     

ToMFIR-based fault detection approach in frequency domain

This paper is to explore further results for total measurable fault information-based residual (ToMFIR) approach to fault detection in dynamic systems. The ToMFIR contains the essential fault information and remains unaffected by control actions in a closed-loop system. It is composed of controller residual and output residual and some of further results are developed in frequency domain. Besides the ability of detecting actuator and sensor faults, it is able to detect faults/failures resulting from the computer used for control purpose that generates control signals. Currently, all of existing fault detection schemes cannot achieve the same task at all. A practical DC motor example, with a PID controller, is used to demonstrate the effectiveness of the ToMFIR-based fault detection. A comparison with the standard observer-based technique is also provided.     

Forecasting time‐varying covariance with a robust Bayesian threshold model

This paper proposes a robust multivariate threshold vector autoregressive model with generalized autoregressive conditional heteroskedasticities and dynamic conditional correlations to describe conditional mean, volatility and correlation asymmetries in financial markets. In addition, the threshold variable for regime switching is formulated as a weighted average of endogenous variables to eliminate excessively subjective belief in the threshold variable decision and to serve as the proxy in deciding which market should be the price leader. The estimation is performed using Markov chain Monte Carlo methods. Furthermore, several meaningful criteria are introduced to assess the forecasting performance in the conditional covariance matrix. The proposed methodology is illustrated using daily S&P500 futures and spot prices. Copyright © 2010 John Wiley & Sons, Ltd.     

Forecasting Forward Defaults with the Discrete‐Time Hazard Model

For predicting forward default probabilities of firms, the discrete‐time forward hazard model (DFHM) is proposed. We derive maximum likelihood estimates for the parameters in DFHM. To improve its predictive power in practice, we also consider an extension of DFHM by replacing its constant coefficients of firm‐specific predictors with smooth functions of macroeconomic variables. The resulting model is called the discrete‐time varying‐coefficient forward hazard model (DVFHM). Through local maximum likelihood analysis, DVFHM is shown to be a reliable and flexible model for forward default prediction. We use real panel datasets to illustrate these two models. Using an expanding rolling window approach, our empirical results confirm that DVFHM has better and more robust out‐of‐sample performance on forward default prediction than DFHM, in the sense of yielding more accurate predicted numbers of defaults and predicted survival times. Thus DVFHM is a useful alternative for studying forward default losses in portfolios. Copyright © 2013 John Wiley & Sons, Ltd.