Overview of all included bankruptcy prediction studies
| Authors . | Evaluated models . | Best model . | Sample size . | Period . | Accuracy . |
|---|---|---|---|---|---|
| Wilson/Sharda (1994) | MDA, NN | NN | 129 | 1975–1982 | 97.50% |
| Lacher et al. (1995) | MDA, NN | NN | 282 | 1970–1989 | 91.50% |
| Alici (1996) | Kohonen network | Kohonen network | 92 | 1987–1992 | N.A. |
| Kim (2005) | NN | NN | 654 | N.A. | 80.00% |
| Huang et al. (2008) | DT, MDA, NN, Adapted NN (incl. ratio calculation of input features) | Adapted NN | 660 | 2004 | 97.87% |
| Tsai/Wu (2008) | NN ensemble | NN ensemble | 690 | N.A. | 97.32% |
| Shi et al. (2009) | DT, NN, NN, Adapted NN (incl. bagging), Nearest Neighbour, SVM, ZeroR | Adapted NN | 1,000 | N.A. | 75.60% |
| Lu et al. (2015) | SVM, Hybrid model (SVM + SPSO) | Hybrid model | 250 | N.A. | 99.21% |
| Antunes et al. (2017) | Gaussian process, LR, SVM | SVM | 2,000 | 2002–2006 | 98.13% |
| Kostopoulos et al. (2017) | BN, NN, RF, SVM | RF | 435 | 2003–2005 | 70.19% |
| Alexandropoulos et al. (2019) | LR, NB, NN | NN | 450 | 2003–2004 | 73.20% |
| Ding et al. (2019) | K-Medians Clustering, SAE | K-Medians Clustering | 97,680 | 1996–2016 | 88.46% |
| Jones/Wang (2019) | LR, TreeNet | TreeNet | 4,922,271 | 2009–2013 | 90.40% |
| Mai et al. (2019) | CNN, Adapted CNN (incl. word embedding), LR, RF, SVM | Adapted CNN | 106,821 | 1994–2014 | 85.60% |
| Rainarli (2019) | DT, LR, NB, Nearest Neighbour, NN, SVM, ZeroR | SVM | 120 | 2008–2014 | 85.83% |
| Cao et al. (2020) | BN, DT, LR, NN, SVM | NN | 1,563,010 | 1961–2018 | 83.72% |
| Jang et al. (2020) | LSTM | LSTM | 1,378 | 1980–2016 | 98.54% |
| Soui et al. (2020) | AB, LD, LR, NN, RF, SAE, SVM, XGBoost | SAE | 10,503 | 2007–2013 | 98.00% |
| Uthayakumar et al. (2020) | Ant colony optimization, LR, NN, RBF, RF | Ant colony optimization | 250 | N.A. | 100.00% |
| Sehgal et al. (2021) | LR, NN, SVM | SVM | 1,957 | 2010–2016 | 83.60% |
| Authors . | Evaluated models . | Best model . | Sample size . | Period . | Accuracy . |
|---|---|---|---|---|---|
| Wilson/Sharda (1994) | MDA, NN | NN | 129 | 1975–1982 | 97.50% |
| Lacher et al. (1995) | MDA, NN | NN | 282 | 1970–1989 | 91.50% |
| Alici (1996) | Kohonen network | Kohonen network | 92 | 1987–1992 | N.A. |
| Kim (2005) | NN | NN | 654 | N.A. | 80.00% |
| Huang et al. (2008) | DT, MDA, NN, Adapted NN (incl. ratio calculation of input features) | Adapted NN | 660 | 2004 | 97.87% |
| Tsai/Wu (2008) | NN ensemble | NN ensemble | 690 | N.A. | 97.32% |
| Shi et al. (2009) | DT, NN, NN, Adapted NN (incl. bagging), Nearest Neighbour, SVM, ZeroR | Adapted NN | 1,000 | N.A. | 75.60% |
| Lu et al. (2015) | SVM, Hybrid model (SVM + SPSO) | Hybrid model | 250 | N.A. | 99.21% |
| Antunes et al. (2017) | Gaussian process, LR, SVM | SVM | 2,000 | 2002–2006 | 98.13% |
| Kostopoulos et al. (2017) | BN, NN, RF, SVM | RF | 435 | 2003–2005 | 70.19% |
| Alexandropoulos et al. (2019) | LR, NB, NN | NN | 450 | 2003–2004 | 73.20% |
| Ding et al. (2019) | K-Medians Clustering, SAE | K-Medians Clustering | 97,680 | 1996–2016 | 88.46% |
| Jones/Wang (2019) | LR, TreeNet | TreeNet | 4,922,271 | 2009–2013 | 90.40% |
| Mai et al. (2019) | CNN, Adapted CNN (incl. word embedding), LR, RF, SVM | Adapted CNN | 106,821 | 1994–2014 | 85.60% |
| Rainarli (2019) | DT, LR, NB, Nearest Neighbour, NN, SVM, ZeroR | SVM | 120 | 2008–2014 | 85.83% |
| Cao et al. (2020) | BN, DT, LR, NN, SVM | NN | 1,563,010 | 1961–2018 | 83.72% |
| Jang et al. (2020) | LSTM | LSTM | 1,378 | 1980–2016 | 98.54% |
| Soui et al. (2020) | AB, LD, LR, NN, RF, SAE, SVM, XGBoost | SAE | 10,503 | 2007–2013 | 98.00% |
| Uthayakumar et al. (2020) | Ant colony optimization, LR, NN, RBF, RF | Ant colony optimization | 250 | N.A. | 100.00% |
| Sehgal et al. (2021) | LR, NN, SVM | SVM | 1,957 | 2010–2016 | 83.60% |
AB = AdaBoost; BN = Bayesian Network; CNN = Convolutional Neural Network; DT = Decision Tree; LD = Linear Discriminant Analysis; LR = Logistic Regression; LSTM = Long Short-Term Memory; MDA = Multivariate Discriminant Analysis; NB = Naïve Bayes; NN = Neural Network; RBF = Radial Basis Function; RF = Random Forest; SAE = Stacked Auto Encoder; SPSO = Switching Particle Swarm Optimization; SVM = Support Vector Machine
Source(s): Table created by author