REXS : A financial risk diagnostic expert system

Artificial intelligence techniques are rapidly emerging as important contributors to more effective management. One of the greatest growth areas probably lies in the use of Expert System methodology for supporting managerial decision processes. Existing Decision Support Systems often attempt to apply analytical techniques in combination with traditional data access and retrieval functions. One of the problems usually encountered while developing such decision support systems is the need to transform an unstructured problem environment into a structured analytical model. Using an expert system approach to strategic decision making in such unstructured problem environments may provide significant advantages. The financial Risk diagnostic EXpert System (REXS) concentrates on Financial Risk Analysis. Based on a Forecasting Model the system will, with the support of several expert system knowledge bases, attempt to evaluate the financial risk of a business and provide guidelines for improvement.

The financial Risk diagnostic EXpert System (REXS) concentrates on Financial Risk Analysis.Based on a Forecasting Model the system will, with the support of several expert system knowledge bases, attempt to evaluate the financial risk of a business and provide guidelines for improvement.

INTRODUCTION
The financial failure of industrial businesses has become one of the most important aspects of any economy.A business is a legal failure if the assets of the business are not sufficient to meet the legally enforceable claims of creditors.It is a proven fact that the underlying cause of most failures can best be summarized by the term "w~nagement incompetence".The usual "causes" given for financial failure, e.g.,lack of capital, faulty accounting, poor planning, etc., are often not causes but rather rationalizations or excuses for poor w~nagement performance.
The two most important factors concerning financial failure are the economic and financial risk of the business.Economic risk is associated mainly with the industry to which the business belongs, and with the general conditions of the particular economy.Financial risk, on the other hand, is subject almost completely to the discretionary control of management.
To help w~nagement in detecting, or even forecasting possible financial failure, a research project, consisting of the development of a Forecasting Model, was carried out at the Bureau of Financial Analysis at the University of Pretoria [9].The model measures the risk of financial failure by means of calculating a risk factor through the combination of ratios determined from items in financial statements.This risk factor is a reflection of the probability that a large business may fail financially.The lower the value of the calculated risk factor, the greater the risk that the business may fail financially.
A financial Risk diagnostic EXpert System (REXS), was developed to further improve the significance and usefulness of the forecasting model, by adding five diagnostic modules to this risk analyses module.The system will enable the user not only to analyse the business, but also to diagnose the problems indicated by the analysis, therefore providing a closed-.loopintegrated .system.
The system consists of several components as follows , a)Data Management , All the data needed by the system is maintained in three different data resources.The business being analysed is classified into a risk class according to a rating determined through time series analysis.

c)Ratio Value Analysis
The ratios that constitute the Financial Forecasting Model are analysed and priorities are allocated according to the importance for financial improvement.

DESIGN STRATEGIES AND OBJECTIVES
The system is based on the [19,20], which supports processing capabilities.expert system development environment GURU a broad range of integrated knowledge The system consists of seven individual, yet integrated, components, each accessed from the main menu.The first component of the system is a data base to control, organize, store, retrieve and update all information needed by the system.The remaining six components of the system use expert system facilities to assist a financial analyst to diagnose the financial risk of a business.The logical procedure may be described as follows : a) Capturing data from financial statements, b) Classification of a business, using user experience, c) Allocating priorities, d) Comparing results with industry trends, e) Simulating improvements recommended by the user, and f) Reviewing the general financial implications.
As previously noted the diagnostic components of on an expert system environment.Conversely, consults various knowledge bases in the course allows the user to directly interact with the the system are based the system itself of its execution and system.When a user presents a particular problem, the system uses the available stored expertise to infer advice which is presented to assist the user when asked for further information.
The specific objectives of the system are as follows : a) Analysis and interpretation of Financial Statements, b) Leading the user through the decision making process, c) Setting certain directives based on user experience, and d) Providing a total review of the financial environment.

Diagnostic data base
In the diagnostic data base information, e.g.user ratings, user priorities, industry ratio values, etc., used in the diagnostic modules of the system are maintained and stored.The detail tablc hart of this data base is shown in Figure 3.

Diagnostic analysis
The diagnostic analysis involves several expert knowledge bases.An expert system is a computer based consultant that has access to stored expertise about some problem domain.When a user presents a particular problem to the expert system, it uses the available expertise to infer sDme advice which is presented tD the user.An expert system has three majDr parts a)StDred expertise : Expertise abDut a prDblem represented as a rule set, which is a cDllectiDn rule has a prem~se and a cDnclusiDn.Whenever engine can establish that a premise is true, then Df the rule is also valid.
domain may be Df rules.Each the inference the conclusion b)Inference engine : An expert system's inference engine is the software that carries out the reasoning needed to solve a prDblem.This software draws upDn the stDred expertise in Drder to reach its conclusions.
c)User interface : When consulting an expert system, the user states a prDblem or gDal and interacts with the system.
One may think of an expert system obtain information, rather than infDrmatiDn algorithmically.
as containing knowledge abDut hDW to as a prDcedure which prDcesses SDme

COMPLETION Results of consultation :
Improvements proposed by the system, and the user are displayed on the screen or printed.The information needed by the different rule sets, are obtained from the three data resources.The inference engine will first use the information stored in the two data bases to reach a conclusion, and then use this advice to lead the user through a series of procedures.

PERFORMANCE OF THE SYSTEM
The performance of the system is demonstrated by means of a practical example.
4.1 Financial statement data management Information in the financial statement data base should be valid and up to date before attempting to execute any of the other functions provided in the system.

CIPS
Figure 5. Sample item value report.

Classification according to risk factor
First the Risk Factor for every year is calculated and reported.
Secondly the classification analysis according to these values as well as ratings supplied by the user is performed as follows

USER CLASSIFICATION RATING
The risk factors are further analysed over time by setting ratings for 12 different classes.Ratings are allocated to 12 different classes by an expert system rule set containing classification criteria, and also by the user.These ratings are set beforehand as a standard to measure the performance of the business over the time period.
This rating is a percentage value.The higher the rating the better the financial growth of the business over the time period analysed.
CLASSIFICATION BASED ON RATING This option will again classify the business, this time using all the information obtained from the previous operation.An expert system is consulted, and a classification report generated, summarizing all the classification information pertaining to the business.

Allocating procedure
The system will consult the corresponding priority allocation knowledge base for each ratio, determining a priority according to the stored information in the rule set.This priority will be shown to the user to assist the user in the decision making process of allocating a priority according to his own judgement.This decision making process is well structured by using priorities to determine the degree of importance of every ratio.The greater the importance of a ratio for improvement the higher should be the priority allocated to this specific ratio.

Revision
This option is used to revise all the priorities allocated to the different ratios.Two graphs are used to display firstly, the ratio values and secondly the allocated priorities.

Comparitive industry and company goal values
Comparative ratio values must be set as evaluation criteria, before any further evaluation may be performed.The following three different evaluation criteria are used in this system : HISTORICAL DATA FROM WITHIN THE BUSINESS : The ratio values of the last three years are calculated.These criteria are used throughout the system.
DATA'FROM OUTSIDE THE BUSINESS: Corresponding data from the industry, in which the business being analysed is operating, is used for calculation purposes.The weighted average of the different ratio values serve as an excellent normative guideline for the ratio values of the business.
A sample output report is shown in Figure 8. COMPANY GOALS : By its very nature financial decision making involves purposeful behaviour, which implies the existence of a company goal, or what is much more likely, some combination of goals.Corresponding company goal values for the different ratios, are set by the management of the company.The output report for these values is similar to the report for the industry values.

Diagnostic simulation
In the preceding modules different classification schemes were used to classify the different financial criteria, in order to prepare the groundwork for improvement application.If this analysis of risk and classification schemes is to be meaningful for the business, a method must be found to translate this analysis and classification schemes into operational decision rules for improvement of the financial risk.

IMPROVEMENT RECOMMENDATIONS
The risk analysis and various classification schemes mentioned, are all mutually inclusive and are all used in the diagnostic sinlulation to recommend improvements and simulate a new financial environment.
A knowledge base containing certain decision criteria based on the risk factor analysis and classification schemes from the preceding modules will determine an improvement value.This improvement value is a value proposed by the system, to be added to the present risk factor value in order to find a better financial risk value.This consultation will be followed by a graphical presentation of the different risk values.This graphical presentation will give the user an idea of the improvement needed by comparing the industry and company goal values with the present risk factor value.The user will be asked to propose an improvement value, using the improvement recommended by the system as a measure.After entering an improvement value a summary of the implication of these improvements proposed by the system and the user will be displayed on the screen.A sample of this screen is shown in Figure 9. PROPOSEO   Each ratio value will be handled individually.The procedure for the ratio values is precisely the same as for the risk factor.

SIMULATION OF A NEW' FINANCIAL ENVIRONMENT
A new financial environment is simulated.This calculation of the improved ratio values, and the the new improved risk factor value.A sample improved values is shown in Figure 10.The user may compare the new simulated risk factor value, determined from the improved ratio values, with the improved risk factor and the present business risk factor value.If the user is not satisfied with the improvement in the risk factor value he can revise the proposed improvements on the different ratio values.
This sensitivity analysis may be repeated until the simulated value compares favourably with the improved risk factor value proposed by the user.
from the REXS by the system, in the form of emanating determined presented

Review diagnostic analysis
This module is used to review the results system.A review of all the improved values and improvements proposed by the user are graphs.
The different graphs are used to display the results obtained from the diagnostic analysis and to compare the improved values with the present values.A sample review graph is shown in Figure 11.A summary screen form containing the different values i.e., values of the last three years, comparative industry and company goal values and improved values emanating from the system, will be displayed on the screen.A sample of this screen form is shown in Figure 12.Planning is the key to the financial manager's success.Financial plans may take many forms, but any good plan mus•t be related to the firm's existing strengths and weaknesses.This indicates that all financial plans must begin with the type of financial analysis presented in the REXS system.If this analysis is to be meaningful for a business, a method must be found to translate this analysis into operational decisions.A diagnostic environment has been developed that allows the user to recommend certain improvements according to personal judgement, and simulate a financial environment based on these improvements.These improvements may be revised for an unlimited number of times, in order to try and find the best alternative as a basis for future financial planning.The system contains a diagnostic simulation module, providing the facility to run different sensitivity analyses, testing the trainee's ski11s in financial risk diagnostics and improvement.

General
The REXS system concentrates on Financial Risk Analysis."Based on a Forecasting Model the system will calculate a financial risk factor reflecting the probability that a business will fail financially.By breaking the risk factor down into the components constituting the risk factor, causes for financial failure are determined through classification, comparison and evaluation.These mutually inclusive analyses are focussed on the improvement of the financial risk factor.A diagnostic simulator is used to simulate a new financial environment, setting guidelines for management to improve the financial risk of the business.
The application possibilities of the system are : a) Ratio analysis over time, b) Classification of businesses into risk classes, c) Specific problem identification, d) Comparison of business growth with industry trends, and e) Setting guidelines for future financial improvement.
The system in general sensitivity analyses, being analysed.provides the facilities to do a number of to improve" the financial risk of the business

Figure 1 .
Figure 1.Conceptual view of the system.

Figure 3 .
Figure 3. Detail table chart for the diagnostic data base.

3. 3
.1 KNOWLEDGE REPRESENTATION Domain knowledge is stored in the fDrm of rule sets.The system cDntains 26 different rule sets, which may be divided intD four grDups : a) Classification rule sets, b) Rating allocation rule sets, c) Priority allocation rule sets and d) Improvement rule sets The ratiD value improvement rule sets are used fDr demonstration purposes.This rule set determines the possible improvement needed on a specific ratio value after comparing theratiD value of the three years analysed, with the ratio values of the industry and the company goal.Figure 4 describes the cDmponents Df these rule set.GOAL IMUSVL -Improvement proposed by user.INITIALIZATION Data obtained from tables: Ratio value of first year, Ratio value of second year, Ratio value of third year, Ratio value of industry, Ratio value of cow~any goal.CONSULTATION Uses comparative values from the industry and the company goal to compare calculated ratio values.and to propose lmprovements according to information in the system, and user expertise.

Figure 4 .
Figure 4. Representation of ratio value improvement rule sets.

Figure 10 .
Figure 10.Sample simulated value form

5. 2
Training The system has the potential of becoming an excellent training tool for business school students and managers.The objectives of such a training program will generally include : a) Better understanding of Financial Risk, b) Better insight into Ratio Analyses, c) Better insight into Comparative Analyses, d) Better insight into managing by objectives, e) Testing of personal judgement, and f) Development of diagnostic skills.
Financial statement data baseIn the financial statement data base all the components, e.g.,items, ratios, etc., needed to calculate the risk factor of the Forecasting Model are maintained and stored.The detail table chart of this data base is shown in Figure2.