Development of Statistical Discriminant Mathematical Programming Model Via Resampling Estimation Techniques

Ziari, Houshmand A.; Leatham, David J.; Ellinger, Paul N.

doi:10.2307/1244291

Cited by 13 publications

(9 citation statements)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Sub-sample jackknife technique was initially proposed by Wu (1990) and extended by Politis and Romano (1994). In the 1990s, researchers including Politis et al (1997), and Ziari et al (1997) used these re-sampling techniques for their empirical work. Compared to standard jackknife, in subsample jackknife more than one observation is dropped to estimate out-of-sample forecasts of the remaining m ¼ nÀd observations, where n is the total number of observations, and d ¼ 2, 3, .…”

Section: Jackknife Out-of-sample Forecastsmentioning

confidence: 99%

Forecast performance of neural networks and business cycle asymmetries

Kiani

Bidarkota

Kastens

2005

Applied Financial Economics Letters

View full text Add to dashboard Cite

Forecast performance of artificial neural network models are investigated using Ashley et al. (1980) and the neural network nonlinearity test proposed by Lee et al. (1993) is employed to find possible existence of business cycle asymmetries in Canada, France, Japan, UK and USA real GDP growth rates. The results show that neural network models are more accurate than linear models for in-sample forecasts. However, when comparing the out-of-sample, linear models performed better than neural network models in all series. Results from neural network tests show that business cycle asymmetries do prevail in all the series.

show abstract

Section: Jackknife Out-of-sample Forecastsmentioning

confidence: 99%

Forecast performance of neural networks and business cycle asymmetries

Kiani

Bidarkota

Kastens

2005

Applied Financial Economics Letters

View full text Add to dashboard Cite

show abstract

“…Future research is needed to compare trained models on completely separate data sets. Although some theories have been presented on sampling methods, these are still in their infancy (14).…”

Section: Discussionmentioning

confidence: 99%

“…Leave-one-out gave similar variance to the other resampling methods but cross-validation. The leave-one-out procedure assures that, regardless of the sample size, the relevant observation would not be in the pseudo-optimal solutions more than once (14). This resampling technique can easily produce significantly different results in NN settings, depending on the training-stopping criterion (15).…”

Section: Discussionmentioning

confidence: 99%

“…For a large sample size and a small B value, bootstrap does not ensure that all the relevant observations will be deleted at least once. The advantage is that in small size samples and with a relatively large B value, bootstrap algorithms may capture variability of sample data better than the leave-one-out procedure (14). Particularly in NN settings, bootstrap yields rather reliable estimates of the variance even in small sample situations, models where the distribution of residual depends on the input and, in addition, it is more robust if the selected model is incorrect (13).…”

Section: Discussionmentioning

confidence: 99%

See 1 more Smart Citation

Validation Procedures in Radiological Diagnostic Models. Neural Network and Logistic Regression

1999

View full text Add to dashboard Cite

The objective of this paper is to compare the performance of two predictive radiological models, logistic regression (LR) and neural network (NN), with five different resampling methods. One hundred and sixty-seven patients with proven calvarial lesions as the only known disease were enrolled. Clinical and CT data were used for LR and NN models. Both models were developed with cross validation, leave-one-out and three different bootstrap algorithms.The final results of each model were compared with error rate and the area under receiver operating characteristic curves (Az). The neural network obtained statistically higher Az than LR with cross validation. The remaining resampling validation methods did not reveal statistically significant differences between LR and NN rules. The neural network classifier performs better than the one based on logistic regression. This advantage is well detected by three-fold cross-validation, but remains unnoticed when leave-one-out or bootstrap algorithms are used.

show abstract

“…• The results obtained by mathematical programming models are easily interpreted. [Ziari et al (1997)]. As Baesens (2003) comments, it is very important to have not just robust and powerful scorecards but also scorecards whose results can be interpreted easily by its users.…”

Section: Introductionmentioning

confidence: 99%

The use of MSD model in credit scoring

Falangis

2007

Oper Res Int J

View full text Add to dashboard Cite

A credit scoring classification problem can be defined as a decision process in which information from application forms for new or extended credit is used to separate the applicants into good and bad credit risks. In the credit industry, it is important to find a method that optimally separates applicants into 'goods' and 'bads' as good classification models can provide competitive advantage. These classification models can be developed by statistical techniques (e.g. statistical discriminant analysis and logistic regression), neural networks and mathematical programming (MP) discriminant analysis methods, although MP methods are less widely used in practice in spite of their advantages, e.g. MP methods are non-parametric and desired classifier characteristics can be represented by constraints in the MP model. In this paper, a MP model is described and compared with other known methods, using real data. The MP model uses minimization of the sum of the deviations of misclassified observations from the discriminant function as its objective function. The performance of this MP model is evaluated on three datasets for credit card applications and is compared with the performance of a k-NN classifier, discriminant analysis, support vector machines and and logistic regression.

show abstract

Development of Statistical Discriminant Mathematical Programming Model Via Resampling Estimation Techniques

Cited by 13 publications

References 13 publications

Forecast performance of neural networks and business cycle asymmetries

Forecast performance of neural networks and business cycle asymmetries

Validation Procedures in Radiological Diagnostic Models. Neural Network and Logistic Regression

The use of MSD model in credit scoring

Contact Info

Product

Resources

About