4.4.2 Logit models

If to compare the results of FGMDH with the results of FNN TSK, one can see that neural network has better accuracy at the short forecasting interval (1 year), while fuzzy GMDH has better accuracy at the greater intervals (2 or more years).

In Table 23, the comparative results of application of different methods for

For estimation of fuzzy methods' efficiency at the problem of bankruptcy risk forecasting the comparison with crisp method, the regression analysis of linear models was performed. As input data, the same indicators were used, which were found optimal for FNN. Additionally, the index net financial result was also included in the input set. This index makes great impact on forecasting results. Thus, input data in these experiments were eight financial indicators of

This conclusion coincides with similar conclusion for Ukrainian banks.

4.4 Application of linear regression and probabilistic models

bankruptcy risk forecasting are presented

Forecasting results of different fuzzy methods.

256 European banks according to their reports:

4.4.1 Regression models

Input data period

Table 22.

Table 23.

Total number of errors

Comparative analysis of forecasting results for FGMDH.

errors

Method (period) Total number of

% of errors

Accounting and Finance - New Perspectives on Banking, Financial Statements and Reporting

2004 7 14 0 7 2005 6 12 1 5 2006 4 8 1 3 2007 2 4 0 2

> % of errors

ANFIS (1 year) 4 8 0 4 TSK (1 year) 1 2 0 1 FGMDH (1 year) 2 4 0 2 ANFIS (2 years) 8 16 4 4 TSK (2 years) 5 10 0 5 FGMDH (2 years) 4 8 1 3

Number of first type errors

Number of first type errors

Number of second type errors

Number of second type errors

debt/assets—X1;

100

loans/deposits—X2;

net interest margin—X3;

ROE (return on equity)—X4;

ROA (return on assets)—X5;

Furthermore, the experiments were performed using logit models for bankruptcy forecasting [9, 10]. The training sample consisted of 165 banks and the testing sample of 50 banks.

The first one was constructed, linear logit model, using all the input variables. It has the following form (estimating and forecasting equations):

$$I\_Y = \mathbf{C}(\mathbf{1}) + \mathbf{C}(\mathbf{2}) \ast X\_1 + \mathbf{C}(\mathbf{3}) \ast X\_2 + \mathbf{C}(\mathbf{4}) \ast X\_3 + \mathbf{C}(\mathbf{5}) \ast X\_4$$

$$+ \mathbf{C}(\mathbf{6}) \ast X\_5 + \mathbf{C}(\mathbf{7}) \ast X\_6 + \mathbf{C}(\mathbf{8}) \ast X\_7 + \mathbf{C}(\mathbf{9}) \ast X\_8$$

$$Y = \mathbf{1} - \text{@CLOIGSTIC}(-\text{(C1)} + \text{C2}(\mathbf{1}) \ast X\_1 + \mathbf{C}(\mathbf{3}) \ast X\_2$$

$$+ \mathbf{C}(\mathbf{4}) \ast X\_3 + \mathbf{C}(\mathbf{5}) \ast X\_4 + \mathbf{C}(\mathbf{6}) \ast X\_5 + \mathbf{C}(\mathbf{7}) \ast X\_6 + \mathbf{C}(\mathbf{8}) \ast X\_7 + \mathbf{C}(\mathbf{9}) \ast X\_8()$$

The next constructed model was a linear probabilistic logit model with six independent variables. The final table including the forecasting results of all the logit models is presented below (Table 25)


#### Table 24.

Comparative analysis of ARMA models.


#### Table 25.

Comparative analysis of logit models.
