Literature Review:

Financial analysis systems are complex, nonlinear, dynamically changing systems, and artificial intelligence is used to create DSS and KMS helping managers and clients to understand complex financial systems. Various financial analysis systems have been developed but due to changing market and globalization such systems has failed to give good results. Due to this, the potential use of neural networks, fuzzy logic and their combination in the form of hybrid intelligent system as a tool for predicting financial markets has been marketed at increasing levels in recent years. Weigand claims that the complexity of a model useful for forecasting may not be related to the actual complexity of the system. Potentially models can accurately predict markets where they are substantially less complex than the market itself. Lowe (1994) focuses on portfolio optimization and short term equity forecasting. Some believe that efficient market theory causes predictions based upon historical price patterns to be valueless. However, Lowe (1994) postulates that A system which is apparently random could have significant deterministic components embedded in its data. He states that a neural networks ability to create nonlinear approximations to the underlying generators of datamay be exploited. Lowe (1994) concludes that it would be possible to develop an automated trading system based entirely upon quantitative pattern processing techniques capable of consistently outperforming professional traders. Lederman and Klein (1995, p. 65) provide Juriks thoughts on trading system development. Although Jurik does not provide specific examples of trading systems, he provides a wealth of advice on data preprocessing techniques. He states, Strive for simple models having only a few choice input variables. increases. He supports this by explaining that as the number of model inputs increase, the degrees of freedom of the governing equation also

Kartalopoulos (1996), Dhar and Stein (1996), an Ward and Sherald (1995) all mention to varying degree that neural networks have the capability to forecast financial markets. Smolensky, Mozer, and Rumelhart (1996, p. 395) provide Weigands thoughts on time series analysis and prediction. Although extremely technical, it touches on financial market prediction and provides a good overview of time series analysis. He breaks time series analysis into forecasting and modeling. Forecasting is short-term prediction while modeling tries to identify features that accurately predict long term trends. Wiegand states that these can be quite different and that the laws governing a short-term forecast may not substantially relate to the long-term model or the actual characteristics of the system. While equations with high degrees of freedom have the capability to model the training data effectively, they fail miserably when given test data. This is because models with fewer degrees of freedom do not try to trace the datas random scattering but only follow the general trend. Jurik also states that When trying to remove unimportant variables, sensitivity analysis of nonstationary or nonlinear models has dubious practical value. This is extremely important because this is one of the standard techniques used in regression analysis. If applied to a neural network model it could seriously fail. Jurik states there are only two ways to correctly remove unimportant variables. The first is to use a genetic algorithm to develop multiple combinations of input variables while only letting the most accurate survive. The second is a manual method of systematically removing one variable at a time and recording network accuracy. This technique is repeated until the accuracy of the model starts decreasing. Toroku, Kumamoto (1996) stressed to buil a financial expert system based on fuzzy theory and Fuzzy Production System (FLOPS), the study consists of four parts. In first part the basic features of expert systems were presented, in second part, fuzzy concepts and the evaluation of traditional expert systems to fuzzy expert systems were presented, for the third part expert

system shell (FLOPS) used in the study were described, for the fourth and last part the development of financial diagnosis system is described, it was developed using Walls seven ratios, traditional seven ratios and also thirty four ratios selected by a financial expert. After analyzing and investigating a financial diagnosis system was developed using a fuzzy expert system which used membership functions based on averages and standard deviation.

Mainly the expert systems have been programmed by expert system tools, the reason of this is because there is lot of flexibility for high level reasoning strategies in these expert system tools. With in many commercial based tools available in the market, FLOPS were selected for the purpose of dealing with ambiguity. In this research paper, the ratios were selected by experts experience, but it is important to select the appropriated ratios using a scientific method, at that time Numeric Analysis to select ratios started. The FLOPS took more time to run because the programs were coded in procedural computer language such as C language. Rimvydas Simutis (2000), build a human skill based fuzzy expert system for decision making support in a stock trading process. The focus was concentrated on computer software that is capable to reproduce the knowledge from the skilled stock trader. The author using classical techniques and soft computing methods developed expert stock trading support system. The system was tested through historical collection of stock market records. The quality of the fuzzy trading system was tested by Paper profit making process for selected stocks using yearly historical records. The original fuzzy system was capable to make about 22% annual profit. For improvements in original fuzzy trading system evolutionary programming techniques was applied. The improved decision support system conceded 30% annual profit.

Tapio, Petri (2001) presented Fuzzy logic-based forecasting model, in their study they presented fuzzy logic-based software tool, fuzzy logic advisory tool (FLAT), for demand forecasting of signal transmission products. The FLAT was developed for the prediction of demand of thousands of different products in order to aid the purchasing process of materials of about fourteen thousand components in the electronic manufacturing process of Nokia Network Systems Haukipudas Factory. The prediction values of different products were deduced by starting from a set of eight input values. Each input value was fuzzied by fuzzy logic advisory tool. The fuzzy results were deduced in three sequential phases, in each phase the number of variables is split due to hierarchical structure of the inference module. The database and rule base were divided into three hierarchical structure of the inference module. Rules are represented by linguistic relations changed into matrix equations from in-order to apply linguistic equations techniques (LE). Fuzzy membership functions for input values were determined by online values of the products. Fuzzy rules were inferred by analyzing behavior of the products together with market experts and product experts of the company. The Fuzzy logic based forecasting model is able to produce more accurate decision making than more traditional approaches. The results obtained from fuzzy logic advisory tool for prediction of demand for signal transmission devices were shown and discussed, design principles for a general predictor was also explained. The tool was developed by combining traditional fuzzy set theory with the linguistic approach, which provides method for developing adaptive expert systems without rule-based programming. The algorithm was implemented as a computer program and applied to real manufacturing data of the company. These kinds of approach were useful for taking account of non-linearity in real-world applications. The potential benefits of a fuzzy logic with linguistic relations seem to be better and more accurate decision making due to model

based approach. From the presented examples in the research paper it was cleared that adaptive membership functions generation with a tree-like hierarchical fuzzy reasoning is a promising tool for manufacturing optimization in the field of high tech engineering business. The system was easily extendable to take into marketing and sales personals expectations from the market developments. In addition, the Fuzzy logic advisory tool can also be used for the educational purposes to illustrate the interactions of different variables. Oscar, Castillo (2001) proposed new method for the estimation of the fractal dimension of geometrical object using fuzzy logic techniques. The fractal dimension is a mathematical concept, which measures the geometric complexity of an object. The algorithms for estimating the fractal dimension calculate a numerical value using as data a time series for the specific problem. The numerical values gives an idea of the complexity of the geometrical object, but there lies a underlying uncertainty in the estimation of the fractal dimension because we use only a sample of points of the object, and also because the numerical algorithms for the fractal dimension are not completely accurate. The authors proposed new definition of the fractal dimension that incorporates the concept of a fuzzy set. The new fuzzy fractal approach was applied to the problem of forecasting the prices of two consumers goods. With a approach proposed by Oscan and Castillo a set of fuzzy rules using fractal dimension of the objects to solve the problems of forecasting can be build. R.J. Kuo, C.H. Chen, Y.C. Hwang (2001) developed a genetic algorithm based fuzzy neural network to formulate the knowledge base of fuzzy inference rules which can measure the qualitative effect on the stock market. The effect was further integrated with the technical indexes through artificial neural networks. The results in the paper indicated that the neural network considering both the qualitative and quantitative factors of the stock market surpasses the neural networks considering only the quantitative factors.

Arnold F. Shapiro (2002) in his approach purposed that, it may have been sub-optimal, in the sense that studies may have been constrained by the limitations of the technology and opportunities may have been missed to take advantage of the synergies between the technologies. For example, NNs have the positive attributes of adaptation and learning, they have the negative attribute of a black box syndrome. Fuzzy Logic has the advantage of approximate reasoning but the disadvantage that it lacks an effective learning capability. Merging the technologies of fuzzy logic, neural networks provide an opportunity to capitalize on their strengths and compensate for their shortcomings. The author presents an overview of the merging of NNs, FL and GAs. The study showed that merged techniques can provide a more accurate and robust solution than can be derived from any single technique. Furthermore, there is relatively little redundancy because typically the methods do not try to solve the same problem in parallel but they do it in a mutually complementary fashion. The NN-based studies suggested avenues to explore that involve the synergies between the technologies. One enhancement would be the use of fuzzy signals and/or weights. Many of the studies used linguistic-type input variables and they could be extended using the fuzzy NNs of Buckley and Hayashi (1994a). Another enhancement opportunity follows from the observation that, while some researchers used GAs to choose NN parameters, such as the learning rate and momentum coefficient, their approach was static. It would be interesting to compare their results with those obtained using a dynamic fuzzy rules approach, along the lines of Bonissone (1998). Only FL-based studies took advantage of the synergies between FL and the other SC technologies so there is lot of opportunity to extend the studies.

Hussein and Pepe (2002), purposed method relies on fuzzy logic to formulate a decision making when certain price movements or certain price formations occur. The success of the system was measured by comparing system output versus stock price movement. The new stock evaluation method was proven to exceed market performance and can be an excellent tool in the technical analysis field. The paper examined fuzzy logic systems in the finance arena. It simulated human behavior in reacting to stock price movement and formation. A number of inputs were created to recommend buy=sell of a specific stock when certain price formation exists. Author simplified some details on investing procedures (cost per transaction is constant, taxation etc.) in order to highlight the properties of the methodology presented. Author applied fuzzy informational technologies to investments through technical analysis. Most investment models are of a proprietary nature and their characteristics are not reported in the scholarly literature. Pruitt and White is a published as-assessment of the potential of technical analysis. They used three technical indicators (cumulative volume, relative strength, and moving average) to build their trading model. Our system examined various companies and proved to be effective. The investment returns were excellent. Most companies consider the performance of S&P 500 as average performance. The results surpassed the S&P 500 by a substantial amount. The buy and sell trigger level can produce different results. According to author analysis, the decision to choose the trigger levels for sell and buy depends on the investor and the stocks long-term trend. Different strategies can be implemented using the fuzzy indicator to match the investor preferences and the industry conditions.

H.S. NG, K.P. LAM , S.S. LAM (2003), designed Genetic Fuzzy Expert Trading System (GFETS) to simulate the vague and fuzzy trading rules and give the

buy-sell signal.

Fuzzy trading rules

are

optimized

and

selected using

genetic algorithm in GFETS. Experimental evaluations showed that trading with the optimized fuzzy trading rules obtains a good profitable return. To maintain the quality of the fuzzy trading rules being in-used, GFETS must be re-trained from time-to-time. In this paper, an incremental training approach was studied and evaluated with all Hang Seng China Enterprises Index (HSCEI) stocks. The risk and the profit return compared with other trading strategies were reported. The profit return including transaction cost for the incremental GFETS was better than that for the common Great China stock funds. Ching-Torng, Ying-Te (2004), in their research explained that Bid/no-bid decision-making is a critical activity for a contractor. Limited by both the nature of and the competition for bid opportunities, bid/no-bid decision is associated with uncertainty and complexity. Because subjective considerations such as nature, competition, value of the bid opportunity, resource capabilities, and the reputation of the company are relevant to the bid/no-bid decision, a fuzzy logic approach is adopted. In this approach assessments are described subjectively in linguistic terms, while screening criteria are weighted by their corresponding importance using fuzzy values. As an illustration, a practical example for evaluating a bid opportunity of an international co-development commercial airplane project in Taiwan was cited, which shows how this method can provide the analyst with more convincing and reliable results and cost saving for the company. To prevent dissipating an organizations energies in preparing losing proposals, a bid/no-bid decision must be made within the context of the organizations strategic framework. Due to ill-dened and ambiguous criteria which exist in bid/no-bid assessments, conventional crisp evaluation approaches cannot suitably and effectively handle such assessments. Thus, a method for bid/no-bid decision-making using fuzzy linguistic was devised.

Results indicate the proposed approach is very useful in a bid/no-bid decision-making process. Although the discussion of this paper is confined to the bid opportunity at ADC Company, the methodology can be applied to other bid opportunities in other companies. The approach proposed in this research paper has several advantages. 1. The method gives the analyst more convincing and reliable results. The fuzzy attractiveness rating is expressed in a range of values. This provides an overall picture about the possible success of bidding and ensures that the bid/no-bid decision is not been biased. 2. The method speeds up the bid/no-bid decision process. It is simple and systematic in terms of evaluation and computation, allows the analyst to evaluate the rating and weight directly using linguistic terms. Fuzzy numbers can easily approximate linguistic terms and compute the fuzzy attractiveness rating of a bid opportunity. 3. The method improved the communication of the rationale behind the committee team and ensures consistency in the decision process. Furthermore, the method provides the company with cost saving (reduction of the man-hours for proposal preparation by about 1525%) and a number of improvements:

Because the initial bid/no-bid decision is speeded up by this method, the proposal team now has more time to prepare the proposal. More time implies that a better proposal can be produced, increasing the likelihood of the customer awarding the company the contract and in this way increasing the companys bid win ratio.

The method aids the company to arrive at an optimal allocation of bidding resources. Less desirable work can be rejected or bid for with reduced effort, affording the company with the resources to address more attractive RFPs.

E.W.T. Ngai and F.K.T. Wat (2005) described the development of a fuzzy decision support system (FDSS) for the assessment of risk in e-commerce (EC) development. A Web-based prototype FDSS is designed and developed to assist EC project managers in identifying potential EC risk factors and the corresponding project risks. A risk analysis model for EC development using a fuzzy set approach was proposed and incorporated into the FDSS. EC development takes place in a complex and dynamic environment that includes high levels of risk and uncertainty. This study has outlined an approach to the assessment of the risks associated with EC development using FST. A model of fuzzy risk analysis was proposed to assist EC project managers and decision makers in formalizing the types of thinking that are required in assessing the current risk environment of their EC development in a more systematic manner than before. A Web-based FDSS was designed and developed to incorporate the proposed risk analysis model. System evaluation was performed to ascertain whether the FDSS achieved its designed purpose, and the results were satisfactory. The feedback and comments collected from respondents were used to make necessary adjustments. The results of the evaluation strongly support the viability of the study approach to risk analysis using fuzzy sets, and demonstrated the feasibility of evaluating EC project risk. The FDSS prototype focused on risk

identification, analysis, and prioritization. Less attention was given to the risk management planning, resolution, and monitoring that is associated with EC development. Further research should be conducted into such risk management planning. In addition, risk monitoring should be conducted regularly to track the status of the identified risks. With such insight and improvement, the FDSS could be further enhanced to handle the functionality of risk management. Moreover, it was assumed that the weighting assigned by each evaluator in the risk evaluation was the same, but the relative importance placed on certain factors by individual decision makers and experts could be widely different.

Zai-En HOU and Fu-Jian DUAN (2009, p. 3) concludes that The philosophy of forecasting nancial index with neural network model has be developed. With this we can approximate some complex functions. For large-scale problem, we can decompose it as several middle-scale or small-scale ones and use the methods to obtain the index. Abdalla Kablan (2010) proposed a new framework for high frequency trading using a fuzzy logic based momentum analysis. Kablan proposed adaptive neuro fuzzy inference system combining the predictive properties of neural network, with the reasoning mechanism of fuzzy logic to create and automated trading and forecasting system. He supported this system with the most suitable model Sugeno Model, which uses if-then rules and each rule is the linear combination of the input variables plus a constant term. Kablan did momentum analysis by categorizing the market into 7 conditions and used current volume to determine the momentum in the market of two companies Vodafone, Nokia. Kablan concluded that further research has to be done to increase the performance and to carter it the use of genetic algorithm is important.

This

research

uses

artificial

intelligence

technique

(fuzzy

logic)

for

developing and implementing a system to predict the financial market. Additionally and potentially most importantly for the investor, model accuracy probabilities will be generated. This will be done by combining financial statement analysis with the Sugeno Model accuracy probability.