Rating System

SOFTWARE CerisRating.exe

The CerisRating is a software able to assign financial rating judgements to firms with scarce information. From the one hand, often financial accounts of firms can present missing values and from the other hand, partnerships are not required to deposit detailed accounts but simplified documents. In particular, in order to evaluate the financial solidity of a firm it is necessary to point out that debts can be towards different subjects, for example providers or banks. Complex systems can help to generalize and forecast rating scores even if there are missing values.

The model proposed is able to analyse and underline the relationships among very simple variables in order to assign financial judgements to firms.

Data used for the elaboration of the software is extracted from the AIDA database that is bought by the IRCrES-CNR (former CNR-Ceris).  

The proposed software is very useful, especially for micro and small entrepreneurs that want know the healthy status of their firm, indeed this information can be strategic for obtaining a financing. In addition, the CerisRating can be interesting also for banks and financial institutions because the software is very simple and easy to use, generating a clear score on financial solidity of enterprise.  


Technical details on artificial neural networks model can be found in the following paper: “Falavigna G., (2012), Financial ratings with scarce information: A neural network approach. Expert Systems with Applications, 39(2), pp. 1784-1792, doi:10.1016/j.eswa.2011.08.074, ISSN: 0957-4174”.

The software has been entirely written with Matlab (R2010a version) and compiled with the appropriate mcc compiler. The obtained file is an executable (*.exe) that can be run on whatever personal computer after the installation of a specific tool (MCR installer) that collect all the necessary libraries.

Clicking on the executable, the software opens a DOS prompt that it must stay open until the software runs.

Figure 1 presents the economic and financial information required for the computation of rating. The amounts have to be expressed in euros. The interface is very friendly: there is the possibility to activate a calculator (i.e., press the button “Calcolatrice”, figure 2), to delete the content of the cells (i.e., press the button “Reset”) and to compute the rating score (i.e., press the button “Calcola il rating”).

The window “Calcolatrice” allows to do computations and also to directly assign the amount to the specific cell.

Before calculating the score, the software makes a control on data correctness, verifying some relations (for example: total assets=total liabilities). If the software recognizes a mistake, a pop-up message appears with the information on the error (figure 3).

Without error, the rating score appears in the cell “RATING” in the starting window interface. Here, not only the rating class is proposed (from D, failed firm, to AAA, very healthy enterprise) but the cell changes also the background colour, based on the legend proposed in the mask (figure 4).  In addition, figure 4 shows that also the size of firm is proposed base on sales (i.e., “micro” if sales < 2 mil€; “media” if 2mil€ < sales < 10 mil€ and “altro” if sales are bigger than 10 mil€). As explained in the paper (Falavigna, 2012), the model for the definition of rating has considered the size. The software selects the firm on the base of size and after it runs the more appropriate framework of the model built on 7 artificial neural networks.

The user can decide to save a file with the information of firm and the rating obtained. With this aim, the button “Report” allows the user to put some company information (figure5) to be written in the report (the output is a file Report.pdf, figure 6). This file is saved in the same folder of the software CerisRating.

Last but not least, after the rating evaluation, the software compares the information of firm to evaluate with those of firms of the same size and rating class. This control informs the user about a possible incorrectness of data and then of the obtained rating. Figure 7 shows that a pop-up alerts about the possibility of error and, at the same time, also stars (***) near to the rating class appear.

Rating Classes

Different letters correspond to different classes of default risk. In particular: