IAES Inter national J our nal of Artificial Intelligence (IJ-AI) V ol. 8, No. 4, December 2019, pp. 391 398 ISSN: 2252-8938, DOI: 10.11591/ijai.v8i4.pp391-398 r 391 Intelligent cr edit scoring system using kno wledge management Bazzi Mehdi 1 , Chamlal Hasna 2 , El Kharr oubi Ahmed 3 , Ouaderhman T ay eb 4 1,2,3 Department of Mathematics and Computer Science, Laboratory of Modeling, Analysis, Control and Statistics, F aculty of Sciences A ˆ ın Chock, Hassan II Uni v ersity , Casablanca, Morocco 4 Department of Mathematics and Computer Science, Research Laboratory in Materials Science and Modeling, soult an moulayslimane uni v ersity , Khouribg a, Morocco Article Inf o Article history: Recei v ed July 25, 2019 Re vised Sep 11, 2019 Accepted Oct 18, 2019 K eyw ords: Credit limits Credit scoring Entrepreneurship Kno wledge management ABSTRA CT Promoting entrepreneurship among Moroccan young people has been challenged by a plethora of economic and social problems in the aftermath of the Arab Spring. Se v eral go v ernment programs ha v e been set up for young entrepreneurs. Thus, f aced with the lar ge number of credit applications solicited by these young entrepreneurs, banks resorted to artificial intelligence techniques. In this respect, this article aims at proposing a decision-making system enabling the bank to automate its credit granting process. It is a tool that allo ws the bank, in the first instance, to select promising projects through a scoring approach adapted to this se gment of entrepreneurs. In the second step, the tool al lo ws the setting of the maximum credit amount to be allocated to the selected project. Finally , based on the kno wl edge of the bank’ s e xperts, the tool proposes a breakdo wn of the amount granted by the bank into se v eral products adapted to the needs of the entrepreneur . Copyright c 2019 Insitute of Advanced Engineeering and Science . All rights r eserved. Corresponding A uthor: Bazzi Mehdi, Department of Mathematics and Computer Science, Laboratory of Modeling, Analysis, Control and Statistics, F aculty of Sciences A ˆ ın Chock, Hassan II Uni v ersity , Casablanca, Morocco. T el: 212 615 10 65 13 Email: bazzimehdi@gmail.com, hasnachamlal@gmail.com 1. INTR ODUCTION The W orld Bank classifies Morocco as a lo wer -middle-income economy [1]. The country is characterized by a gr eat potential for gro wth thanks to a relati v ely de v eloped economic di v ersification compared to other countries in the Middle East and North Africa (MEN A) re gion. Despite the global financial crisis and political unrest in the re gion stirred by the Arab Spring, the Moroccan economy [2] achie v ed a gro wth of 3% in 2018. According to Doing B u s iness 2017, Morocco ranks second in Africa and fourth in the MEN A re gion [3]. In se v en years, the country has g ained 60 ranks in this ranking mo ving from 128th position in 2010 to 68th in 2017. In addition, in 2018, the Micro, small (MSEs) and medium-sized enterprises (SMEs) represent 99% of the national economic structure (94% of MSEs and 5% of SMEs). This f act prompted the Moroccan go v ernment to adopt se v eral measures to impro v e the b usiness climate for young micro-entrepreneurs, which is the subject of this study . F or the purpose, this paper pro vides a set of re gulatory measures (statutes specific to MSEs, auto-entrepreneurs, compan y re gulations), administrati v e measures (reduce formalities and processes for the creation of companies), and tax measures (tax e x emptions and benefits for ne wly created companies), without for getting support before and after the creation and the financing arrangements. J ournal homepage: http://ijai.iaescor e .com Evaluation Warning : The document was created with Spire.PDF for Python.
392 r ISSN: 2252-8938 Banks, o n their side, pro vide support for these young entrepreneurs to realize their projects. Thus, f aced with the mass of credit requests solicited by these young entrepreneurs, banks are encouraged to put in place tools to fluid the procedures for handling these applications. In this conte xt, we propose a decision- making tool bas ed on the kno wledge management [4] of the bank’ s e xperts to select promi sing projects and propose financing formulas adapted to their needs. At first, we will ta ckle the concept of Kno wledge Management (KM) and its v arious f acets in banking. Secondly , we will present the process follo wed to acquire the kno wledge of the e xperts used in the s election of promising projects and the formulas used for their finances. Finally , we will apply to a credit application all the kno wledge acquired from b usiness e xperts. 2. KNO WLEDGE MAN A GEMENT IN THE B ANKING SECT OR In the f ace of an e v er more stringent re gulatory en vironment combined with strong commercial competition and the rene w al of their b usiness, banks are determined to put in place measures to promote the management of their or g anizational kno wledge. Indeed, according to [5] in their article ”Financial risk and the need for superior kno wledge management”, t h e transfer of kno wledge to decision mak ers beforehand, a v ailable access to information, or the generation of ne w kno wledge on the e v olution of risk management requirements, should lead to more ef ficient risk management. According to [6], kno wledge management is the management of kno wledge within an enterprise through specified or g anizational procedures for the acquisition, or g anization, maintenance, application, sharing and updating of emplo yees’ kno wledge to impro v e their performance and create v alue. In their article ”What is kno wledge management for banks”, the authors [7] propose three possible readings of kno wledge management: its finality , its place of production, and its discipline. Dif ferent f acets of kno wledge management as sho wn in Figure 1. Figure 1. Dif ferent f acets of kno wledge management The first axis deals with the finality of kno wledge management. Indeed, tw o dif ferent streams are required. The first, in a logic of creation of the ne w kno wledge to inno v ate [8]. The second in a replication logic of good practice [9]. Banks that ha v e adopted the kno wledge management approach ha v e put in place systems for kno wledge sharing (acti vity reports, procedure manual) based on Lotus notes, intranet and internet, a customer relationship management system [10] or softw are allo wing emplo yees access to the digital library (e learning) to de v elop their kno wledge of the banking profession. The second axis of analysis is the place of production of kno wledge. In f act, tw o places of conception of kno wledge oppose one another . The first is e xternal [11] adv ocated by the a u t hors who c ombine customer relationship management with kno wledge management. It is a question of kno wing the profiles of the bank’ s customers in order to of fer those customized products and services. The second place is internal [12] and IJ AI, V ol. 8, No. 4, December 2019 : 391 398 Evaluation Warning : The document was created with Spire.PDF for Python.
IJ AI ISSN: 2252-8938 r 393 concerns the kno wledge mobilized in or g anizational practices. It in v olv es initiating a process of e xchange and brainstorming between emplo yees to create and transfer kno wledge within the bank. The third axis of analysis identifie s the users of the concept of kno wledge management within the bank. In f act, tw o major department ha v e been identified. It is primarily the information systems department [13] whose objecti v e is to put in place a system that allo ws the process of creation, collecti on, or g anization, access and use of kno wledge to be automated as much as possible. The second department that uses kno wledge management is the human resources . The latter is interested in the management of job forecasts and the creation of a f a v orable conte xt for learning and kno wledge sharing within the bank [14]. In the follo wing, we aim to pro vide risk management function within a bank with a credit decision support system whose purpose is to replicate good practices in the selection and financing of young entrepreneurs. 3. FUNCTION AL AN AL YSIS OF THE CREDIT DECISION SUPPOR T SYSTEM (CDSS) The CDSS system (which we propose) is a decision support system which, follo wing the capitalizat ion of the e xpert’ s reasoning, will allo w the selection of the good projects of the young entrepreneurs, generate credit lines, their amounth and the adequate guarantees. The tool, as sho wn in Figure 2, will be able to combine se v eral quantitati v e criteria (model for the assignment of scores and calculation of credit limits) and qualitati v e criteria deduced from the reasoning of b usiness e xperts follo wing se v eral meetings. The system is a response to se v eral needs from a set of potential users (in bold): Figure 2. Description of the credit decision support system a. First, the system will allo w the b usiness entity ; In particular , account managers gi v e concise, ef ficient and reflecti v e guidance to customers by focusing more specifically on credit lines that are appropriate to their needs. b . The system directly serv es the entr epr eneur , who will not ha v e to w ait for ef fecti v e advice b ut will be serv ed in a v ery precise and timely manner , which will allo w him/her to sa v e time and e v en stronger confidence in the services of the bank. c. The error is human, and this is one of the main constrai nts that the system will a v oid to an y risk analyst , since the tool will impro v e the risk management of an y application by being more automatic and follo wing a well-chained process. The entity in char ge of the Risks will be able thanks to this tool to retrace all the stages on which it based itself to gi v e its decision concerning the credit limits and the guarantees solicited. d. The system can also serv e the Risk Management entity by a v oiding forcing and by f acilitating the writing of the appropriate reports to each application thus allo wing a f acility of risk management. e. V oluntary departures, resignations or retirements can influence the proper approach of the bank’ s services and gi v e rise to a great loss of kno wledge and e xperience. The bank will be forced to train new r ecruits ne w recruits and thus lose the kno wledge that has been b uilt up o v er time with great e xperience. The Intellig ent cr edit scoring system using knowledg e mana g ement (Bazzi Mehdi) Evaluation Warning : The document was created with Spire.PDF for Python.
394 r ISSN: 2252-8938 system, in this case, will be able to capi talize and unite this reasoning to become a pedagogical tool, which wil l mak e it possible to identify a well-b uilt reflection and a v oid the loss of the e xperiences of former collaborators. 4. THE ARCHITECTURE OF THE CDSS SYSTEM 4.1. Description Figure 3 sho ws the architecture of the credi t decision support system (CDSS) allo ws us to outline the o v erall design of the system by describing its essential components. The CDSS system will recei v e in the first place all the inputs in particular , the kno wledge of the b usiness e xpert resulting from the intervie ws and meetings as well as the data subtracted from the internal system of t he bank (Project data, en vironment, Profile of the project holder ...). The data cited will serv e as a kno wledge base (f acts and rules) to an inference engine that serv es as a receptacle and allo ws to map the reasoning mechanism, then to ha v e a graphical interf ace that the non-e xpert and e xpert user will use to display the v arious proposals for e xits to kno w (lines of credit, their limits and the appropriate guarantees). The system has a tw ofold meaning since the non-e xpert user can inte grate the credit request at the graphical user interf ace (GUI) that will be recei v ed by the inference engine to feed the system’ s f act base. Figure 3. Description of the credit decision support system 4.2. The CDSS kno wledge base The CDSS system is based on a kno wledge base that brings together in a structured w ay the rules, relationships and problem-solving strate gies for the credit application. The content of the CDSS kno wledge base w as acquired through a series of meetings with b usiness e xperts, through intervie ws and structured meetings. The course of these meetings generally consisted of identifying each rele v ant element and asking the e xpert to produce the rules used daily . The kno wledge base is di vided into tw o components, a f act base, and a rule base. 4.2.1. The fact base The f act base is a database that records all e xisting historical data and information in the bank’ s internal system, which will e xploit the bank to deri v e a concise structure of quantitati v e and qualitati v e data [15]. In our article [16]” we ha v e de v eloped a qualitati v e model to e v aluate the projects of young entrepreneurs. The authors, after hundreds of hours of discussion with se v eral stak eholders (risk management e x ecuti v es, netw ork manager , etc.) and dealing with se v eral e v aluation models, ha v e identified 10 to 20 questions from a starting list containing 100 and about 200. The selected questi o ns co v er tw o important aspects of the select ion of young entrepreneurs’ projects. This is a questionnaire designed to assess the ability of the young entrepreneur to manage the project and a n ot her questionnaire to assess its feasibility . The final e v aluation of the project is the combination of the ”Entrepreneur” and ”project” scores. The follo wing equation gi v e the final score: F inal s cor e = E ntr epr eneur scor e (62%) + P r oj ectscor e (38%) (1) IJ AI, V ol. 8, No. 4, December 2019 : 391 398 Evaluation Warning : The document was created with Spire.PDF for Python.
IJ AI ISSN: 2252-8938 r 395 The authors justify the o v erweighting of the entrepreneur block by the f act that a good e n t repreneur can succeed i n an a v erage project, whil e a less trained (or less supervised) entrepreneur can f ail e v en with a promising project. The rating scale v aries between 0 and 100, a score of 100 is gi v en to the best couple (Entrepreneur , project). Finally , the final score is se gmented into 7 classes as sho wn in T able 1. T able 1. Model MasterScale Classe Probability of def ault A 2,10% B 3,90% C 3,93% D 6,70% E 10,00% F 14,40% G 17,70% H 35,10% I 46,20% The criterion used to split the score into se v eral classes is to group in one class projects with the s ame risk profile and to disperse those with dif ferent risk profiles. Once the projects are selected, the question of their financing arises. In the follo wing, we will present the rules and models used to set the credit limits granted to young entrepreneurs carrying selected projects, as well as the breakdo wn of this limit by type of product (Cash, discount, consumption, foreign e xchange). 4.2.2. The rule base a. Setting the cr edit limit: In our article ”Concentration risk: setting credit limits in loan portfolios, case of Morocco” [17], we propose an analytical method which will allo w us to decide for a contractor carrying the project selected, the amount of credit limit to be granted to it according to its risk profile and the risk appetite of the bank about this entrepreneur and his project. The model is as follo ws: CC e = PD e limit e + ( V ar (99%) E L ) limit e E with: CCe: The consumption of capital allocated to this entrepreneur; PDe: The probability of def ault; V ar (99%): V alue at risk that presents the maximum loss that the credit institution is lik ely to incur by financing selected projects with a probability of 1% o v er a one-year horizon. EL: Expected loss (to be pro visioned) associated with the funded project portfolio. E : T otal amount to finance the selected projects. In what follo ws, we will present the rules of e xperts used to allocate the authorized limit to a project leader in a set of products (Consumption, e xchange . . . ) adapted to his needs. b . Br eakdo wn of the global limit: The kno wledge base of the e xpert and his e xperience will allo w our system thanks to the set of criter ia deduced to guide us to w ards the choice of lines of credit. The b usiness rules and the kno w-ho w of the e xpert will mak e it possible to set and calibrate these lines of credit to get closer to the authorization or the limit that the credit institution must grant to the cus tomer . In our system, as an illustration after the intervie ws and the analysis of the data recei v ed from the b usiness e xperts, we ha v e been able to implement some criteria allo wing the choice of lines of credit as sho wn in the T able 2. Intellig ent cr edit scoring system using knowledg e mana g ement (Bazzi Mehdi) Evaluation Warning : The document was created with Spire.PDF for Python.
396 r ISSN: 2252-8938 T able 2. Extract from the rule base (1/2) Acti vity of the entrepreneur Need of the entrepreneur Import: Customs Economic Re gime T emporary admission Import: do wn payment method and deals A v al local supplier Import Export Lines of e xchange Sector Construction and Public W orks CPro visional / final security and Adv ance on Mark et Discount P ayment method: commercial bills Escompte Seasonal acti vity Credit companion The table abo v e reflects the reasoning of the b usiness e xpert when choosing or directing a contractor for a line of credit. At the le v el of this table we appl y a schematization of the reasoning of the human brain; (If, then. . . ). These b usi n e ss rules, which we will e xplain in the T able 3, allo w the credit analyst to deduct the amounts that the credit institution should authorize for each product line. T able 3. Extract from the rule base (2/2) Credit lignes limit amounth guarantees Pro visional Guarantee =T urno v er soumissions bank guarantee / definiti v e rate payment deadlines /360 AS Mark et =T urno v er administration rate payment deadlines /360 Escompt = T urno v er payment mode rate commercial bills payment deadlines /360 is in itself a guarantee Cash f acilities 1 mounth of the turno v errr the pledging of a b usiness b usiness the pledging of treasury bills and go v ernment bonds Collaterals F actoring Untel 90% of the bill personnal g arantee proof of the operation : bills AS Commodities Untel 80% the commodities pledge agreement Insurance Credi Con v ention Openning F ore x e xchange line Untel 6 Mois of international turno v erl 5. EXPERIMENT A TION 5.1. Description of the case As an illustration of the application of our decision tool, we propose the follo wing case: a. An entrepreneur operating in the construction sector , who forecasts a turno v er of 12 000 000 MAD and who hopes to carry out international operations (Import-Export). b . The compan y is paid up to 20% by and o v er a period of 60 days, up to 50% by transfer and 30% by endorsement. c. On the other hand, the compan y pays its clients up to 20%, 30% by ef fects o v er a period of 60 days and 50% by ef fects o v er a period of 90 days. d. The compan y carries out a seasonal acti vity and follo ws a customs economic re gime and bids in a public mark et up to 70%, the share of the administration is 3%. e. The compan y plans an international turno v er of MAD 1,000,000. After inte grating the pre vious data into our system, the results are as follo ws: 5.2. Choice of lines and cr edit limits fixed by the expert T aking into account the rules specified abo v e, an analysis of the credit application independently on the risk profile (mesured by probability of def ault [18]) of the client and the risk appetite of the bank to w ards this client allo ws the latter the follo wing the set of products, as sho wn in T able 4. IJ AI, V ol. 8, No. 4, December 2019 : 391 398 Evaluation Warning : The document was created with Spire.PDF for Python.
IJ AI ISSN: 2252-8938 r 397 T able 4. Credit limits by product Proposed products Limits (En MAD) % Limits (En MAD) without risk adjustment with risk adjustment T emporary admission, IT Documentary remittance documentary credit cash discount 2 100 000 57% 1 640 838,21 Credit companion pro visional g arantee 27 300 1% 21 330,90 definiti v e g arantee 27 300 1% 21 330,90 cash adv ance to get commodities 39 000 1% 30 472,71 fore x e xchange line 500 000 14% 390 675,76 cash f acilities 1 000 000 27% 781 351,53 T otal amounth autorised 3 193 600 100% 2 886 000 Ho we v er , the analysis of the risk profile of the client through the credit scoring model [16] a nd the model of setting the credit limits [17] allo ws us the follo wing results, as sho wn in T able 5. T able 5. Risk profile of the client and its etimated limit Entreprise classe Autorisation Capital limit estimated Gap (limit estimated, Autorisation) SME C 3 193 600 200 000 2 886 000 -28% The credit limit calculation model authorizes this client an amount equal to 2.886.000 MAD. Therefore, we need to do wngrade the amount allo wed by the e xperts (a decrease of 28%). One w ay to re vie w this amount is to apply the shares of each product to the risk-adjusted authorized amount. Indeed, the authorized amount for the product cash dis count for e xample is 2 100 000 MAD, or 57% of the initial amount (without risk adjustment). An appl ication of this percentage to the theoretical li mit of 2 886 000 MAD mak es it possible to de v ote to this product an amount of 1.897.733 MAD. F or this credit application, the proposed CDSS system slightly lo wered the amount allo wed to the selected project applicant because of its a v erage risk profile. Ho we v er , for other profile considered less risk y by the system, the limit is re vised upw ards and the same approach is applied to find the appropriate set of product to the application studied. 6. CONCLUSION The go v ernment’ s ambition to promote the b usiness climate has enabled the deplo yment of se v eral programs to support the economic inte gration of young entrepreneurs. Ho we v er , successful program rollout deserv es strong bank in v olv ement through the funding of promising projects. Hence the need for these credit institutions to set up systems, based on artificial intelligence, t hat can deal with the issues of young entrepreneurs in a timely manner and also support these young people. Thus we ha v e established a decision- making tool (SADC) allo wing the bank, selection and funding of promising projects based on implicit kno wledge including rating models and design of credit lines, and also on e xpert kno wledge of the banking profession. REFERENCES [1] W orld Bank Country and Lending Groups,2017-2018. [2] Central bank of Morocco, financial stability report, 2018. [3] Doing Business 2017: Equal Opportunity for All [4] Mohamed Bayad, Ser ge Simen, Le management des connaissances: ´ etat des lieux et perspecti v es, 2003 [5] Marshall, C., Prusak, L. and Shpilber g, D. “Financial risk and the need for superior kno wledge management”, California Management Re vie w (38:3), 1996 [6] Ala vi, M. Leidner , D.E. (1998). Kno wledge Management and Kno wledge Management Systems: Conceptual F oundations and an Agenda for Research. [7] V al ´ erie P allas-saltiel et Rania Labaki , Quel management des connaissances pour les ´ etablissements bancaires?, Re vue franc ¸ aise de gestion, 2009, pages 139 ` a 151. Intellig ent cr edit scoring system using knowledg e mana g ement (Bazzi Mehdi) Evaluation Warning : The document was created with Spire.PDF for Python.
398 r ISSN: 2252-8938 [8] NON AKA, I. and T AKEUCHI, H. (1995), The Kno wledge-Creating Compan y , Oxford, Oxford Uni v ersity Press. [9] P ark, H., Ribiere, V ., Schulte, W . D., Jr . (2004). Critic al attrib utes of or g anizational culture that promote kno wledge management technology implementation success. Journal of Kno wledge Management, 8(3), 106–117. [10] Saif and al. An Expert System with Neural Netw ork and Decision T ree for Predicting Audit Opinions, International Journal of Artificial Intelligence (IJ-AI), V ol 2, No 4, December 2013. [11] Chen, M.Y . and Chen, A.P . (2006). Kno wledge management performance e v aluation: A decade re vie w from 1995 to 2004. Journal of Information Science, 32 (1), 17-38. [12] Ar gote, L., McEvily , B., Reag ans, R. (2003). Managing kno wledge in or g anizations: An inte grati v e frame w ork and re vie w of emer ging themes. Management Science, 49(4), pp 571-583. [13] Chalmeta, R., Grangel, R. (2008). Methodology for the implementation of kno wledge management systems. Journal of the American Society for Information Science and T echnology , 59(5), 742-755. [14] Horwitz, F ., Heng, C. T ., Quazi, H. A. (2003, V ol 13 No 4). Finders, k eepers? Attracting, moti v ating and retaining kno wledge w ork ers. Human Resource Management Journal, pp. 23-44. [15] Jalil and al. Modeling with ontologies design patterns: credit scorecard as a case study”, Indonesian Journal of Electrical Engineering and Computer Science, V ol.17, No.1, January 2020, pp429-439. [16] B AZZI and al., Credit Scoring in t h e service of entrepreneurship in Morocco: pragmatic Approach for the selection of promising projects, the Risk Go v ernance and Control Journal, A vril 2016. [17] B AZZI and al., Risk concentration: Setting credit limits in loan portfolios, Case of Morocco, Risk Go v ernance and Control journal, Juin 2016. [18] Arup guha, Prediction of Bankruptc y using Big Data Analytics based on Fuzzy c-means Algorithm, International Journal of Artificial Intelligence (IJ-AI), V ol 8, No 2, June 2019. IJ AI, V ol. 8, No. 4, December 2019 : 391 398 Evaluation Warning : The document was created with Spire.PDF for Python.