USING NEUTRAL NETWORKS TO DETERMINE THE FINANCIAL PLAN

392.Introduction

Financial planning has several principles. Among them is the fact that it follows an enterprise strategy, mission and goals, it follows an acceptable finance policy and the goal it follows is the maximization of enterprise market value (Marek, 2006). In addition, planning is a collective activity (not only the financial manager takes part), while the quality of financial plan is being assessed according to its influence on the development of enterprise market value (García, Bedoya a Ríos, 2010). According to Umansky (2004) financial planning is needed because investment and financial decision-making are related, they influence each other and because of that they can not be done independently. The interesting point is that until recentlythe enterprises did not pay the needed efforts to planning. The reason was the persistent negative attitude due to state economy planning (negative experience with planning in the past), instability of conditions within the business, and insufficient knowledge of the techniques and methods of financial planning (Marek, 2006).

The plan output are, according to Kislingerová (2007) financial statements (planning balance sheet, profit and loss sheet and cash flow plan sheet), which are compiled for the whole planned period and are elaborated in individual accounting periods and further even in individual months. Most authors indicate the division of financial plans in the following way. Strategic plan (5 – 20 years), a long-term plan (1-5 years) and a short-term plan (up to 1 year).

The goal of the strategic and long-term plans is, according to Eschenbach (2000), besides compiling a plan balance sheet, an income statement and a cash-flow balance sheet, it is especially guessing the future sales volume, creating the plan of investment activities and a plan of long-term financial sources. These goals have both financial and non-financial character and are meant rather universally than in value (Garía, Bedoya and Ríos, 2010). On the contrary, the goal of short-term plans is the assets, capital and profit state planning for a certain period, the management of liquidity or revenue and costs (Eschenbach, 2000).

Nowadays we recognise three main methods of financial planning – intuitive, statistical and causal (Baldacci, Boschetti, Christofides,N. and Christofides,S., 2009). THe intuitive method is based only on experience and subjective guessings of the person creating a financial plan, while remissioning causal relations, and that only goes on in the person´s head (Gansel, 2008). The disadvantage of this method is, according to Marek (2006), the simplification and a high probability of omitting significant mutual relations. The result may prove an unrealistic plan.

The meaning of statistical method is the extension of time series in future (Baldacci, Boschetti, Christofides, N., and Christofides, S., 2009). It includes especially the method of regression analysis or proportional property growth, liabilities growh and costs in relation to planned sales (Kislingerová, 2007). The weak point of this method is, according to Marek (2006), an unreal presumption that in-the-past-developing economic variables will stay the same in the future. Better results may be often brought by intuitive and causal methods.

Causal method is understood to be the most optimal possible method. The input data is based on the information about the current enterprise property and the current economy results, on the output and other economic plans, while the source is the prediction of macroeconomic indicators´ development (Kislingerová, 2007). Variables express the desirable values of a part of the indicators in the area of costs, liquidity and asset turnover ratio. Other variables in the planned form of financial statements, that are calculated through a certain formula in which the input or wanted variables are contained in the output variables (Marek, 2006). Control is important as well, checking whether the economy output values, calculated in the planned statement respond to the value in planned sheet. This act is secured by control variables (Baldacci, Boschetti, Christofides, N. and Christofides, S., 2009).

In the past, the intuitive method was in the forefront, which is by far insufficient nowadays, and so other methods of financial planning and predictions of enterprise sufficiency have started to develop. It is the discriminant analyses, regression analyses, time series methods and artificial neural networks, in the first place (Kislingerová, 2007). Prediction of enterprise efficiency should use both financial and non-financial indicators (Joshi and Lam, 2006).

Gansel (2008) states that enterpreneurs should understand the financial range of their decisions, and therefore he suggests a certain framework of financial planning which deals with decision-making, revealing of information and corporate strategy based on an exact financial plan. According to Umansky (2004) the current business environment is very complicated, and often unfavorable. Due to that it is important to focus oneself on strategic planning of a financial situation of an enterprise.

The non-artificial neural network system is an efficient method for solving a number of economic classficiation and regression problems and it represents a modern, widely used computational tool especially where it is impossible or too difficult to use traditional approaches. (Zhang, Z. a Zhang, C., 2002). Artificial neural networks´ task is to replace human thinking which, let´s face it, doesn´t always need to be able to take in and interpret a huge amount of information (Slavici, Mnerie a Kosutic, 2012).

During the process of financial planning are neural networks useful because they are able to learn and having learned they can express both - latent and strongly non-linear dependencies (Zhang, Z. and Zhang, C., 2002). According to Slavici, Mnerie and Kosutic (2012) the partial disadvantage may be the impossibility to guess the error size and state the reliability intervals. Wang, Stockton and Baguley (2010) claim that the success of an enterprise is, up to a point, dependent on the exact prediction of a future development. According to Russel (2011), the application of neuroscience onto financial planning is interesting, but relatively problematic. To be useful in practice it demands a structurally and practically applicable framework. There are several methods for compiling financial plans nowadays, neural networks are, however, partially unappreciated in this sense, because they are used rarely, and financial managers probably cannot fully appreciate their benefits (Zhang, Z. and Zhang, C., 2002). Truth is that a few of the traditional methods are applicable and the problem of neural networks is relatively difficult to be applicable even more (especially for smaller enterprises). It contains a wide range of construction alternatives thanks to which it may be often difficult to choose the right network for a specific requirement (Wang, Stockton a Baguley, 2010).Nevertheless, as Mengel and Wouters have found out, a compilation of a financial plan is very beneficial even for beginning enterprises. A problem rises when their owners claim that they are too busy for developing a planning strategy or they do not have the financial plans in a written form (Marek, 2006). These persons should take part in a dialogue with a qualified financial planner who will certainly confirm the importance of financial plan compilation (Umansky, 2004). Besides artificial neural networks, other intelligent techniques, such as different expert systems, fuzzy logics or genetic algorythms, are employed. (Zhang, Z. a Zhang, C., 2002).Artificial neural networks are used nowadays, in the enterprise sphere especially for predicting future development of the enterprise, the costs development or stating whether the enterprise is or will be bankrupt or creditworthy. According to Wang, Stockton and Baguley (2010), they can ensure a better source planning, monitoring, coordination, but also management as a whole. For the financial plan of an enterprise, compiled according to the neural network methodics, the data and indicators of the whole financial analysis spectrum are important, while often this information is gathered separately and not only is the access to it very easy (Umansky, 2004).

Numbers and money is the language of business. If an individual wants to be successful in running a business, they need to learn this language. The key factor, however, is to understand it and know how to use it (Davenport, 2015).

The goal of this contribution is to outline the possibility of use of artificial neural networks compiling a short-term financial plan on the basis of an exemplary given enterprise.

393.1 Methodology

For the purpose of this text an enterprise´s financial plan will be compiled for 2015. With the help of artificial neural networks (specifically time series) the starting variable values which are significant, will be set. It will be either the statements´ totals to which we will add their partial values or, on the contrary, it will be significant partial statement items, which will be added consequently, to reach the form of totals.

In the case of balance sheet the values for 2015 will be set with the help of neural networks for the following variables: Total Assets (to check the sum), Long-term Assets, Long-term intengible assets, Long-term tengible assets, Long-term financial assets, Current assets, Stock, Long-term claims, Short-term claims, Financial assets, Other assets, Accruals, Total liabilities, Equity, Basic capital, Reserve fund and other fund created from profit, Profit/lost of past periods, Profit/lost of current period, Debt, Reserves, Long-term obligations, Short-term obligations, Other liabilities, Accruals. For Profit and Loss Planning sheet the following items will be used: Sales of own products and services, Material and energy, Services, Salaries, Bonuses to board and memebers of cooperatives, Costs of social security and health insurance, Social Expenses, Taxes and fees, Depreciation of fixed assets, Revenues from sales of fixed assets, Net fixed assets and raw material price sale, Change in reserves and provisions relating to operating activities and complex deferred expenses, Other operating revenues, Other operating costs, Interest income, Interest expenses, Other financial expenses, Income tax payable, Income tax deferred. In case of Cash-Flow we issue through the case of the Indirect Measurement Method from the change of balance-sheet accounts and Costs-and-Profit correction. Therefore only the following items will be used: Net operating cash flow, Net cash flow from investments, Net financial cash flow.

For the calculation of time series the DELL Statistica software will be used, specifically the Neural Networks Module. The data will be transferred from MS Excell file into Statistica. It is possible to calculate neural networks for individual variables, for individual statements or for all financial statements at once. With regard to the range of the contribution it is suitable to calculate all variables using one calculation and so determine an acceptable (demonstrative) number of networks reached within the circuit.The first line of imported file will be regarded as a description of values in columns. Further, the Intelligent Problem Solver tool will be used. Time series will be set. All variables used for the calculation are continuous. Years will be given as the independent variable. We examine the wanted variables of individual items in Financial statement as the dependent variables.

A software will generate random 1000 artificial neural ones, out of which 5 most suitable will be kept. If the error in each additional generated network increases it is possible to terminate the calculation earlier than 1000 artificial neural structures are reached. For the calculation, the following neural networks will be used:

• 
  Lienar,
  
• 
  GRNN (Generalized Regression Neural Network),
  
• 
  Radial Basis Function (further RBF),
  
• 
  Three Layer Perceptron,
  
• 
  Four Layer Perceptron.
  

Each iteration will be made of, as needed, 1 to 10 steps. Models will distribute the results based on the linear function or logit function. Networks characteristics will be calculated only for the resulting best five. Others will not be taken into account.

The results will be interpreted for the whole package, always divided into the training, validation and testing set of data.

Consequently it is necessary to insert the individual results for 2015 into the Reports of Financial Statements and calculate other variables. In case of balance sheet it is necessary to calculate the partial items. That will be marked, based on the part of partial item from 2004 – 2014, on the calculating item. In case of Profit and Loss Statement that will be a combination – i.e. if we know the totals, the division will happen analogically, the way it would in a balance sheet. In other cases we will know all the partial items and subsequently be able to add them up in one single total. In cash flow statement it is possible to gain individual items based on the knowledge of balance sheet and Profit and Loss Statement. Calculated individual cash flow will serve as control values.

394.2 Results

2.1 Neural Networks Neuronové sítě

On the basis of the analysis conducted five best neural networks have been obtained. Particular characteristics are given in table no. 1.

Tab. No. : Five best Neural Networks

Profile	Train Perf.	Select Perf.	Test Perf.	Train Error	Select Error	Test Error	Training/ Members	Inputs	Hidden(1)	Hidden(2)
MLP s4 1:4-100-46:46	0,00	0,00	0,00	0,000000	0,00	0,00	BP28b	1	100	0
Linear s7 1:7-46:46	0,00	0,00	0,00	0,000000	0,00	0,00	PI	1	0	0
MLP s5 1:5-100-100-46:46	0,00	0,00	0,00	0,000000	0,00	0,00	BP2b	1	100	100
GRNN s5 1:5-1-47-46:46	0,00	0,00	0,00	0,000000	0,00	0,00	SS	1	1	47
RBF s5 1:5-1-46:46	0,00	0,00	0,00	0,000000	0,00	0,00	KM,KN,PI	1	1	0

Source: Own

All networks show zero error in all sets. That may be given by the fact that the time series is relatively short. Nevertheless, according to other characteristics, such as prediction analysis, sensitivity analysis or residue analysis, we are able to find out which network is the most suitable one. With regard to the goal of this contribution it is not necessary to analyse each result in detail and deal with the choice of the most suitable neural structure. Picture No.1 represents the schemes of the neural structures.

Pic. No. :Schemes of acquired neural networks







Source: own

6.2 Financial Plan

Based on the sensitivity analysis, prediction power and residue analysis especially, the simplest network, Linear s7 1:7-46:46, seems to bet he most suitable one, in regard of its structure.

That is why it will be continuously used for plan compilation for 2015. Table No.2. offers a shortened balance sheet of CEZ obnovitelne zdroje s r.o. company for 2015.

Tab. No. : Shortened balance sheet of CEZ obnovitelne zdroje s.r.o. company in 2011 – 2015.

Item		2011	2012	2013	2014	2015
	TOTAL ASSETS	4 389 358	389 972	497 999	433 917	372 947
A.	CLAIMS FOR OWN SUBSCRIBED CAPITAL	0	0	0	0	0
B.	LONGTERM PROPERTY	4 232 116	173 045	168 331	161 662	138 947
B.I.	Longterm intangible property	396	493	341	90	77
B.II.	Longterm tangible property	4 109 172	45 004	40 442	33 628	28 903
B.III.	Longterm financial property	122 548	127 548	127 548	127 944	109 966
C.	CURRENT ASSETS	156 221	216 847	329 609	272 129	233 892
C.I.	Stock Zásoby	823	692	287	5	4
C.II.	Longterm Claims Dlouhodobé pohledávky	0	3 400	17 907	44 270	38 050
C.III.	Shortterm Claims Krátkodobé pohledávky	155 373	212 740	311 415	227 854	195 838
C.IV.	Financial property Finanční majetek	25	15	0	0	0
D.	OTHER ASSETS – temporary asset accounts	1 021	80	59	126	108
D.I.	Accruals	1 021	80	59	126	108
	TOTAL LIABILITIES PASIVA CELKEM	4 389 358	389 972	497 999	433 917	372 947
A.	OWN CAPITALVLASTNÍ KAPITÁL	2 319 608	263 864	249 887	200 499	172 327
A.I.	Basic Capital Základní kapitál	1 404 353	118 000	118 000	118 000	101 420
A.II.	Capital Funds Kapitálové fondy	0	0	0	0	0
A.III.	Reserve Funds and other Profit Funds	29 790	33 725	33 725	0	0
A.IV.	Economic Results of the Past Years	806 766	99 589	112 139	33 887	29 125
A.V.	Economic Result of Current Reporting Period (+/-)	78 699	12 550	-13 977	48 612	41 781
B.	FOREIGN SOURCES	2 060 274	126 105	248 112	233 411	200 614
B.I.	Provisions	44 239	18 780	96 770	91 240	78 420
B.II.	Longterm liabilities Dlouhodobé závazky	1 928 356	0	0	0	0
B.III.	Shortterm liabilities Krátkodobé závazky	87 679	107 325	151 342	142 171	122 194
B.IV.	Bank loans and assistance Bankovní úvěry a výpomoci	0	0	0	0	0
C.	OTHER LIABILITIES - temporary liability accounts	9 476	3	0	7	6
C.I.	Accruals	9 476	3	0	7	6

Source: 2011 – 2014 Albertina, 2015 author

The 2011 – 2014 period was inserted to demonstrate the trend and the-at-first-sight- logical corectness of the prediction. The same is true for the shortened version of Profit and Loss Statement for 2014 – 2015, which is the subject of Table no. 3. 2015 is given in a financial plan again.

Tab. No. : Shortened Version of Profit and Loss Statement of CEZ obnovitelne zdroje s.r.o. in 2011 to 2015.

Item		2011	2012	2013	2014	2015
I.	Revenues from sale of goods Tržby za prodej zboží	13	21	15	18	0
A.	Costs incurred for the sale of goods Náklady vynaložené na prodané zboží	5	15		6	0
II.	Performances	813 525	888 366	2 172 638	2 095 865	1 804 772
B.	Power Consumption	137 437	678 075	1 627 826	1 865 082	1 603 016
C.	Personal Costs	97 043	95 470	86 922	39 923	34 313
D.	Tax and Fees	79 032	92 309	396 436	140 616	120 858
E.	Depreciation of longterm intangible and tangible property	258 177	7 447	4 163	2 219	1 907
III.	Revenues from longterm property and material sale	2080	1187	1240	8432	7 247
F.	Amortized cost from the sale of longterm property and material	2 996	685	539	5 142	4 419
G.	Change in Provisions and corrective items in the operating area and cmplex costs in future periods	3 322	-32	85 035	-7 464	-6 415
IV.	Other operating revenue	1 245	21 716	760	476	409
H.	Other operating costs	4 473	19 419	612	3 111	2674
V.	Transfer of operating revenue
I.	Transfer of operating Costs
*	ECONOMY OPERATING RESULT	234 378	17 902	-26 880	56 156	51 656
VI.	Revenues from the sale of securities and shares
J.	Securities and shares sold
VII.	Revenues from longterm financial property	0	0	3 946	3 946	0
VIII.	Revenues from shortterm financial property
K.	Costs from financial property
IX.	Revenues from CP and derivatives revaluation
L.	Revenues from CP derivatives revaluation
M.	Change in provisions and corrective items the the financial area
X.	Revenue Interest	161	351	27	65	56
N.	Cost Interest	116 084	3	56	52	45
XI.	Other financial Revenue	1
O.	Other financial cost	82	131	173	88	76
XII.	Transfer of Financial Revenue
P.	Transfer of Financial Cost
*	ECONOMY FINANCIAL RESULT	-116 004	217	3 744	3 871	-64
Q.	Regular Activity Income Tax	37 165	3 619	-4 699	11 415	9 811
**	REGULAR ACTIVITY ECONOMY RESULT	81 209	14 500	-18 437	48 612	41 781

Source: Rok 2011 – 2014 Albertina, rok 2015 autor

Table No. 4 submits individual items of cash flow during the period of 2011 – 2015, where 2015 is a prediction again.

Tab. No. : Aggregated cash flow of CEZ obnovitelne zdroje s.r.o. Company in 2011 - 2015A

	2011	2012	2013	2014	2015
Net operating cash flow	473 835	-51 926	133 574	-81 111	-69 714
Net cash flow from investments	-120 345	-5 776	-1 075	7 022	6 035
Net financial cash flow	-354 557	57 692	-143 526	74 089	63 679

Source: Rok 2011 – 2014 Albertina, 2015 author

395.Conclusion

The goal of this contribution was to outline the possibility to use artificial neural networks while compiling a short-term financial plan based on an example of a specific company. The goal has been fulfilled. A financial plan for the CEZ obnovitelne zdroje s.r.o. company has been set, based on the application of neural networks, wpecifically it was a linear neural network called Linear s7 1:7-46:46. The result proves that neural network method is applicable. Time series respect the development of previous years (even when there were fluctuations during the individual years). Nevertheless, neural networks do not respect the plans and intentions of the company´s management. I tis then suitable to incorporate such a financial plan within the set of companywide plans, i.e. marketing, production, ensurance of capital goods, etc. To make it reflect the investment intentions of the given company. The method is then limited, however:

1. 
   Neural networks are applicable in an environment of a stable company (not necessary to be growing) .
   
2. 
   Through neural networks the financial manager will manage to prepare the first variation of financial plan.
   
3. 
   Financial plan, which will develop through neural networks, is necessary to be rectified based on the management´s intentions – i.e. the company strategy, and incorporate it within the structure of other company plans.
   

Observing the deviations from the real fulfillment of a financial plan within the next few years always occurs to bet he next research problem. Even such residues may help rectify and re-train another suitable neural network for the upcoming years.

396.References

Baldacci, R., Boschetti, M., Christofides, N., & Christofides, S. (2009). Exact methods for large-scale multi-period financial planning problems. Computational Management Science, 6(3), pp. 281-306.

Davenport, L. (2015). Business Financial Planning For 2015: Four Steps To Success. General Interest Periodicals--United States, pp. 1.

Eschenbach, R. (2000). Controlling. 1st ed. Praha: Codex, 812 p.

Gansel, B. (2008). Financial planning in business venturing. International Journal of Entrepreneurship and Innovation Management, 8(4), pp. 436-450.

García, J., Bedoya, L., & Ríos, C. (2010). Importance of Financial Planning in Business Plan Elaboration and its Impact on Enterprise Development. Revista Facultad de Ciencias Económicas: Investigación y Reflexión, 18(1), pp. 179-194.

Joshi, A., & Lam, S. S. Y. (2006). Business Performance Prediction Using Recurrent Neural Networks. IIE Annual Conference. Proceedings: Department of Systems Science and Industrial Engineering. United States: Institute of Industrial Engineers-Publisher, pp. 1-6.

Kislingerová, E. (2007). Manažerské finance. 2nd ed. Praha: C. H. Beck, 745 p.

Marek, P. (2006). Studijní průvodce financemi podniku. 1st ed. Praha: Ekopress, 624 p.

Mengel, S., & Wouters, M. (2015). Financial planning and control in very small start-up companies: antecedents and effects on company performance. International Journal of Entrepreneurship and Small Business, 26(2), pp. 191-216.

Russell, N. (2011). A Applying Neuroscience to Financial Planning Practice: A Framework and Review. Journal of Personal Finance, 10(2), pp. 10-65.

Slavici, T., Mnerie, D., & Kosutic, S. (2012). Some applications artificial neural networks in agricultural management. Actual Tasks on Agricultural Engineering: Proceedings of the 40.: International Symposium on Agricultural Engineering, Opatija, Croatia, 21-24 February 2012, 40(1), pp. 363-373.

Umansky, R. (2004). Diagnostic financial planning for small business owner(s). Daily Commercial News and Construction Record, 77(129), 5,8. Wang, Q., Stockton, D. J., & Baguley, P. (2010). Process cost modelling using neural networks. International Journal of Production Research [online], 38(16), pp. 3811-3821.

Wang, Q., Stockton, D. J., & Baguley, P. (2010). Process cost modelling using neural networks. International Journal of Production Research [online], 38(16), pp. 3811-3821.

Zhang, Z., & Zhang, C. (2002). An Agent-Based Hybrid Intelligent System for Financial Investment Planning. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), 2417, pp. 355-364.

Contact

doc. Ing. Marek Vochozka, MBA, Ph.D.

Vysoká škola technická a ekonomická v Českých Budějovicích

Okružní 517/10, 37001 České Budějovice, Czech Republic

vochozka@mail.vstecb.cz

Directory: wp-content -> uploads -> 2015
2015 -> Police Militarization
2015 -> 2015 ihbb championships: hs history Bowl Round 4 – Prelims First Quarter
2015 -> Wrtp/big step · 3841 W. Wisconsin Ave., Milwaukee, wi 53208
2015 -> Rao bulletin 1 December 2015 html edition this bulletin contains the following articles
2015 -> Darton College
2015 -> Training Workshop on Planning and Implementation of Multi-Sectoral Development Programme (MsDP) in Bihar Date : 6th
2015 -> Rao bulletin 15 June 2015 html edition this bulletin contains the following articles
2015 -> Rao bulletin 1 September 2015 html edition this bulletin contains the following articles
2015 -> Rao bulletin 1 August 2015 html edition this bulletin contains the following articles
2015 -> Science, and transportation united states senate

Download 2.73 Mb.

Share with your friends: