AN ENTERPRISE FINANCIAL STATE ESTIMATION BASED ON DATA MINING

Transcription

1 AN ENTERPRISE FINANCIAL STATE ESTIMATION BASED ON DATA MINING Mikhail D. Godlevsky, Sergey V. Orekhov Naional Technical Universiy Kharkov Polyechnic Insiue Frunze sr. 2 Ukraine-6002 Kharkov god_asu@kpi.kharkov.ua, osv@kpi.kharkov.ua Absrac: This paper is concerned wih he daa mining applicaion o business process auomaion. A shor uorial on sages of financial analysis of an enerprise is given. A brief descripion of daa mining componens for financial analysis of an enerprise is presened in he paper. The main elemen of daa mining applicaion is fuzzy neural sysem. The srucure of fuzzy neural sysem includes linguisic variable financial sae, neural nework and fuzzy rules. The fuzzy neural sysem forms he financial sae esimaion based on balance shees from daaware house of an enerprise. The daa mining applicaion has been realized as sofware in Microsof Visual Basic 6.0. The sofware was applied o esimaion of financial sae of Kharkov (Ukrainian) enerprises in 999. Finally, he se of enerprises for invesmen was found. Keywords: daa mining, business process, financial sae, fuzzy neural sysem.. Inroducion The business process auomaion of an enerprise is based on modern informaion echnologies. An enerprise is a complex business sysem, which includes main and suppor business processes (BP) of differen ypes and appoinmens. The complex sysem includes he following BP: BP of producion, BP of planning and managemen, BP of ransformaion. Each BP characerizes operaive informaion, which describes is curren sae. The operaive informaion abou business sysem sores in daa warehouse. The operaive informaion is srucured subjecively. Using ools, he company manager analyzes financial sae and direcions of his developmen. Such principle is a basis of daa mining echnology (Fig..). The financial resources managemen BP based on daa mining is invesigaed in he paper. The fuzzy neural sysem (FNS) is used as a ool for operaive informaion processing. FNS esimaes he business sysem financial sae. The financial sae esimaion is he main sage of financial analysis. The mehod of financial analysis includes he following blocks: general financial sae esimaion, invesigaion of esimaion aleraion during year, financial sabiliy analysis, solvency analysis, abiliy o pay analysis, business aciviy invesigaion. 229

2 Each sage of financial analysis uses he special enlismen of financial indexes. The index values are he basis of decision making for business sysem financial aciviy improvemen. Operaive informaion abou business sysem Manager Daa Warehouse Fuzzy Neural Sysem Resuls Business View Fig... Concepual scheme of daa mining using The financial sae indexes divide in wo groups: assessmen indexes and coordinaion indexes. Tradiionally, he main idea of financial analysis is comparison of curren index values wih opimal values, which characerize financially sable business sysem (enerprise) [SHSA96]. Expers formulae he esimaion of financially sable enerprise. Thus, he financial analysis is based on searching dependencies among differen financial indexes and wih he following comparison of index curren values wih opimal values. In his case FNS creaes he nonlinear dependencies among differen classes of financial indexes, which describe financial aciviy of he business sysem [GO98]. 2. Fuzzy neural sysem of financial sae esimaion ρ Noe, he vecor of financial sae indexes, as P = { P,..., P r } : P i = ϕ i ( x,..., x n ), i =, r, where ϕ i ( x,..., x n ) he ransformaion funcion of balance shee absolue parameers x,..., x n ino financial sae index - P; i r quaniy of financial sae indexes; 230

3 ρ X = ( x,..., x n ) - he vecor of balance shee absolue parameers; n general quaniy of balance shee absolue parameers, which i is necessary o use for evaluaion of financial sae indexes from vecor P ρ. The indexes from vecor P ρ are used for he following linguisic variable value definiion: F= {" financial sae", T, U, Gr, Sem}, where T = { T, T2, T 3 } erms of linguisic variable F; U universal se; Gr grammaical rule; Sem semanic rule. The erms of linguisic variable Financial sae are he following noions: T - sable financial sae, T 2 - unsable financial sae, T 3 - crisis financial sae. The values of linguisic variable F characerize fuzzy variables v, v2, v 3, which belong o universal se U. Grammaical rule Gr is he simple enumeraion of erms from se T. The semanic rule is a neural nework. Neural nework evaluaes he values of membership funcions of fuzzy variables v, v2, v 3. The values received are analyzed by fuzzy rules, which define concree erm of linguisic variable Financial sae. The rapezoidal membership funcion was used for evaluaion of fuzzy variables v, v2, v 3, which have he following view [P98]: vi = { y U, Ai ( y)}, i =, 3, where y - linear combinaion of balance shee absolue parameers x,..., x n ; A i ( y) - membership funcion of fuzzy variable. The scheme of FNS is presened on Fig. 2.. The inpus of FNS are he balance shee absolue parameers possessed by vecor X ρ. Inpu signals of FNS characerize balance shee, which was creaed on a concree dae. Usually he dae is he firs number of monh. The oupus of FNS are values of membership funcions of fuzzy variables, v v, which describe he erms of v 2, 3 linguisic variable F. Fuzzy rules analyze he values of membership funcions A i ( y ), i =,3 (see Fig. 2.). Fuzzy rules define he erm-value of linguisic variable F. The fuzzy rule formulaion is based on neural nework oupu values, which correspond o fuzzy variables v, v2, v 3 and have he following srucure [GO98], [P98]: 23

4 ρ IF ( Ai ( y( X )) ρ = NNi ( X ) z i ) THEN Ti F=, i =, 3. Parameer z defines minimal value of membership funcion of fuzzy variable i when linguisic variable F= Ti, i =, 3. Thus, he decision making abou financial sae is being performed in a fuzzy space. v i, W il W hl W ol X N N 2 A ( y ) X 2 2 N 2 2 N 2 Rule T Rule T 2 3 ( A y ) X n n N m 2 N 3 Rule T 3 Inpu Hide Oupu Layer Layer Layer Neural nework Fuzzy rules Fig. 2.. The srucure of fuzzy neural sysem of financial sae esimaion Using special learning algorihm, he weigh marixes of neural nework ( W il, W hl, W ol ) have been evaluaed. 3. Sofware for financial sae esimaion The funcional srucure of sofware for financial sae esimaion is presened in Fig. 3.. The IDEF0-diagram is demonsraed in Fig. 3.. The daa mining applicaion was realized as sofware. The sofware was performed in Visual Basic 6.0. The main form view is presened in Fig

5 The daabase was performed in Microsof Access using DAO echnology. The daabase includes he following ables: business sysem descripion, balance shee, acives, passives. The sofware also checks he reliabiliy of balance shees. The sofware also analyzes he enerprise groups. This sofware faciliy is needed for invesmen analysis. Balance shees Bookkeeping law documens Value of financial sae esimaion Operaional informaion preparaion Fuzzy neural sysem Esimaion of financial sae Financial sae of an enerprise Daabase Reliable values of financial indexes Visual Basic program and sysem sofware Fig. 3.. Funcional srucure of sofware Fig Sofware main form. 233

6 All sofware work resuls are presened in graphical form. 4. The resuls of sofware using for financial sae esimaion of Kharkov enerprises The resuls of financial sae esimaion of Kharkov (Ukrainian) enerprises, which have been received, are presened in Fig. 4.. The operaive informaion for esimaion (see Fig. 4.) included balance shees of fory Kharkov enerprises. Each enerprise had wo balance shees for esimaion: ) balance shee on December, 999; 2) balance shee on December, The Fig. 4. demonsraes how many balance shees were esimaed as sable, unsable and crisis. However, general quaniy of balance shees presened in Fig. 4. is only seveny-nine. This fac means ha one balance shee was esimaed worse han crisis sae. This balance shee characerizes he financial sae of bankrup enerprise. Using he approach proposed, he financially sable enerprises were found. These enerprises were recommended for invesmen in The resuls of financial sae esimaion of he Kharkov enerprise Bolshevik are presened in Fig The esimaion dynamic is presened in Fig This enerprise produces winer laher clohes. In his case he enerprise had more financial resources before and afer winer season. The enerprise had finances before winer season, as hey were needed for producion preparaion. The enerprise had finances afer winer season, because i was profi. This fac is demonsraed in Fig. 4.2 in and posiions. During summer and winer he enerprise did no have enough finances, because, firs, summer period is no a season for laher clohes, second, in winer he enerprise has o direc all finances in producion aciviy Sable Qauniy of balance shees Unsable Crisis 0 0 Terms of lingusic variable Fig. 4.. Resuls of financial sae esimaion of fory Kharkov (Ukrainian) enerprises in

7 63% 70 56% 60 5% Unsable Crisis % 27% Fig Bolshevik financial sae esimaion in 999 (Kharkov, Ukraine). Therefore, analyzing Fig. 4.2 he manager of Bolshevik enerprise has received business view abou annual financial sae. This view he can use for improvemen of business process of financial resource managemen a he nex year. Thus, he approach of financial sae esimaion has been presened in he paper. The approach is based on daa mining. The daa mining applicaion includes he fuzzy neural sysem of financial sae esimaion. Using he applicaion realized, manager of an enerprise is conrolling a business process developmen. Bibliography [GO98] [P98] [SHSA96] Goloskokov A.E., Orekhov S.V. The Task Of Sae Esimaion Of Social Economic Objecs. // Vesnik of Kharkov Sae Polyechnic Universiy, Number 0 - Kharkov pp Pedrycz Wiold. Compuaional Inelligence. An Inroducion. New York: CRC Press p. Shereme A.D., Sayfulin R.S. Mehodology Of Financial Analysis Of An Enerprise. Moscow: Infra-M pp