Neuro Fuzzy based Stock Market Prediction System

Transcription

1 Neuro Fuzzy based Stock Market Prediction System M. Gunasekaran, S. Anitha, S. Kavipriya, Asst Professor, Dept of MCA, III MCA, Dept Of MCA, III MCA, Dept of MCA, Park College of Engg& tech, Park College of Engg & tech, Park College of Engg & tech Coimbatore. Coimbatore. Coimbatore. Abstract - Neural networks have been used for forecasting purposes for some years now. Often arises the problem of a black-box approach, i.e. after having trained neural networks to a particular problem, it is almost impossible to analyze them for how they work. Fuzzy Neuronal Networks allow adding rules to neural networks. This avoids the black-box-problem. Additionally they are supposed to have a higher prediction precision in unlike situations. Applying artificial neural network, genetic algorithm and fuzzy logic for the stock market prediction has attracted much attention recently, which has better correlated the nonquantitative factors with the stock market performance. However these approaches perform less satisfactorily due to the memoryless nature of the stock market performance. In this paper, we propose a data compression-based portfolio prediction model hybridized with the fuzzy logic and genetic algorithm. In the model, the quantifiable microeconomic stock data are first optimized through the genetic algorithms to generate the most effective microeconomic data in relation to the stock market performance. Key words Neural Network Fuzzy logic portfolio Stock Prediction System. 1. INTRODUCTION A number of studies have examined the differences between neural networks and other approaches for modeling and forecasting time series. Especially the question if no linear models like neural networks can beat linear models have been an issue for several years. Nevertheless the solution seems still to be unclear when taking into account the cost - i.e. the computational and methodological expenses. In this paper two types of neural networks are examined. The first is the classical neural network approach where a neural network is used to predict the future price of an asset from the history of the time series itself. The second approach is a family of quite new neural network models where fuzzy logic is combined with neural technology to archive higher precision in forecasting and some additional issues. Fuzzy neural networks provide the possibility to implement rules into neural network topology; a methodic framework is described in practice fuzzy neural networks compete with classical neural networks where no extra information is given than the patterns created out of the time series. An introduction into applying neural networks for casting financial time series is given. The application of rules allows to model patterns which occur not very often and are therefore not very likely to be modeled by a classical neural network. Rules or groups of rules can be modeled which are specialized on specific situations and circumstances In this paper, a new data compression-based portfolio prediction model hybridized with genetic algorithm and fuzzy logic is developed. Traditionally, difference models are applied in the area of portfolio prediction. Two classical views on the prediction are namely, technical and quantitative. The technical view of markets is that the prices are driven by investor sentiment and that the underlying sequence of prices can be captured and predicated well using charting techniques. This method studies the action of the market as a dynamical entity, rather than studying the actual goods in which the market operator. This is a science of recording the historical market data, such as prices of stocks and the volume traded, and attempting to predict the future from the past performance of the Organized by Department Of ECE, ANNA UNIVERSITY COIMBATORE. 354 Electronic copy available at:

2 function of the underlying security valuation, but also governed by investor sentiment, health of the economy and many others. membership functions and FAMs, a fuzzy logic system is finally achieved. Fuzzy logic has been applied very successfully in many areas where conventional model based approaches are difficult or not cost-effective to implement. However, as system complexity increases, reliable fuzzy rules and membership functions used to describe the system behavior are difficult to determine. Furthermore, due to the dynamic nature of economic and financial applications. Rules and membership functions must be adaptive to the changing environment in order to continue useful. This article outlines a Stock Market Prediction System (SMPS) system that extends the neural networks approach to handle fuzzy, probabilistic and Boolean information. An SPS combines the various advantages of expert systems, artificial neural systems and fuzzy reasoninig.it is designed as integrated networks architecture, based on a building block called neural gate. II. LITERATURE REVIEW The Key issues in building the system are knowledge acquisition, processing of Boolean fuzzy and probabilistic data, and integration of expert knowledge with investor preferences, fuzzy reasoning and automatic learning. Recently, the artificial neural networks (ANNs) have been a popularly researched issue. An artificial neural network is a useful tool in enhancing quality of decision-making. ANNs are used in business and banking applications for decisionmaking, forecasting and analysis. In the beginning of the training process, training pattems are fed fiom both sides of the system. Input samples are fed into the input fuzzy layer, and corresponding target samples are fed into the output fuzzy layer. The sample data are to form the initial fuzzy sets in input fuzzy layer and output fuzzy layer separately. The thresholds of neurons in the first and third layers do not use a sigmoid function, but represent fuzzy sets. The last.layer s function is centroid defuzzification. This design means that a fuzzy logic system is hidden in the neural network structure. By using the network training process to obtain fuzzy Fuzzy logic represents a technology for designing sophisticated control systems. The rules to be used for equity fund investment decision-making could take into account the following variables: economic conditions, Relative Strength Index(RSI) and the gap from the moving average (MA gap). Fuzzy inference takes the states of input variables (imprecise linguistic terms such as good or bad ) and constructs a fuzzy trading rule such as IF economic conditions are bad AND the RSI is high, AND the MA gap is high, THEN the trading action is strong sell. Fuzzy system contains a fuzzifier, which converts inputs (or independent variables) into fuzzy variables, and a defuzzifier, which converts the output (or the dependent variable) of a fuzzy control process into numerical value output. In a fuzzy system the process of generating the output (control) begins with taking and fuzzifying the inputs, and then executing the entire active IF THEN-type rules from a predetermined collection of rules designed t o capture the reasoning process, a socalled rule base. The very first step in building an SMPS is the acquisition of expert knowledge and the representation of that knowledge. The knowledge acquisition system gives rise to the rule base illustrated in figure B, which consists of companybased rules, attributes-based rules and country-based rules. The relationship between price and related indexes can be represented by a polynomial function such that Z = XiYi (1) Z = Closing price of trust; X = coefficient: Y = Closing price of quantitative factors (Investment indexes include: - NSE National Stock Exchange - BSE Bombay Stock Exchange) Organized by Department Of ECE, ANNA UNIVERSITY COIMBATORE. 355 Electronic copy available at:

3 The objective for the SMPS is to find the effective coefficients so the optimized price P can be recalculated based on the following formula. III. DEVELOPMENT OF STOCK MARKET PREDICTION SYSTEM (SMPS) As stated before, simply measuring the performance of an ANN by looking at its accuracy on forecasting does not bring useful information about how it would perform in the real stock market, helping the investor to make decisions. A better way to assess the performance of the ANN is to use its outputs as inputs to a trading system. The day-trading system introduced in this section is responsible for "translating" the ANN predictions into business decisions, i.e. when to buy or sell stocks. FIgure 1 Stock Market Prediction System Fig 1. represents the integrated network architecture of an SPS designed to carry out the stock selection function for international investment management. parts: A trading system is composed of three main (i) A set of rules to enter and exit trades, (ii) A risk control mechanism, and (iii) A Portfolio management scheme. Besides, it has to take into account all the constraints imposed by the real market, such as brokerage commission rates, slippage, the volume being negotiated and round lot trades. The proposed trading system works by following the stock market in real time, but takes into account price changes occurred in fixed intervals of 15 minutes (one minute intervals could have been used, but we chose a more conservative approach). Hence, every 15 minutes the system consults its trade rules, and can advice the investor to perform an enter or exit trade. A. Entry and Exit rules for trades Defining the exact conditions in which an investor will choose to buy or sell a stock is the most important part of any trading system. In our system, these rules will tell how the minimum and maximum daily prices provided by the ANN will be used to make decisions about the right time to trade. Hence, every 15 minutes the trading system checks the closing price (for the interval) and compares it with the minimum and maximum predicted prices. If the closing price is smaller than the minimum predicted, the system advices the investor to buy the stocks. If the closing price is greater than the maximum predicted, the investor is advised to sell the stocks. It is important to notice that more than one buy or sell operation is allowed in the same day, and that the order in which they occur is irrelevant (as long as they are alternated). A day can also be closed without any trade. This occurs when the stock price remains higher than the minimum and lower than the maximum predicted values during the whole day. Organized by Department Of ECE, ANNA UNIVERSITY COIMBATORE. 356

4 In contrast, if the first operation is performed (regardless of being a buy or a sell), the second operation is compulsory within the same day. Hence, if at the end of the day only the first operation (entry) was executed but the second operation (exit) was not, it is executed in the last minute of the day, according to the day closing price. Finally, the concept of stop-loss, which is detailed in Section IV- B, is also incorporated into the rules, and work as a risk controlling mechanism. In summary, let minann and maxann be the minimum and maximum values predicted by the ANN, and close be the 15-minute interval closing stock price. are: The 6 rules of the proposed trading system 1) Buy when close ~ minann, sell when close 2:: maxann; 2) Buy when close ~ minann, sell in the last minute of the day; 3) Buy when close ~ minann, sell with the stop-loss price; 4) Sell when close 2:: maxann, buy when close ~ minann; 5) Sell when close 2:: maxann, buy in the last minute of the day; 6) Sell when close 2:: maxann, buy with the stoploss price. B. Risk Control Establishing ways to control the risk involved in the trades is an important feature of a trading system. When we talk about risk, we refer to the amount of money that can be lost in an trade. Usually, a trading system will lead the investor to perform a higher number of trades that will bring him/her loss than trades that will bring him/her gain. The challenge is to make small losses in unsuccessful trades and high gains in profitable ones In this work, the stop-loss strategy is used for this purpose. The stop-loss strategy defines a default price that, if reached, will cause the exit. This price is calculated using the trade entry price. We performed experiments varying the percentage of stop-loss from 0.1% to 2.0%. A low stop-loss can avoid profitable trades, while a high one can cause a higher loss. The system performed better with stoploss rules than without them, and the value of 0.5% gave us the best results. C. Portfolio management Portfolio management refers to the amount of available resources that will be used in each trade, considering the risks involved and the total capital available. As we are using the stop-loss approach, we created a simple methodology where all the available capital is used in all the trades. Different strategies were tested, but they did not give as good results as the ones following this approach. Evaluation Metrics By using a trading system, we can define new metrics to evaluate the ANN performance, based on the results of the business actions taken by following the network "advices". Here we present three evaluation metrics, which will be later used to assess the proposed day-trading system. The first of these metrics is the annualized (percentage) return, defined as AR == 100 x ((FC/IC)(365.25/D) - 1), where IC represents the initial capital invested, FC the final capital obtained and D is the number of days of investment (i.e., it adjusts the return to an annual basis). The second metric is the maximum drawdown. A drawdown represents the total percentage loss of capital experienced by the system before it starts winning again. The maximum drawdown is the highest drawdown occurred during the period considered, and represents a way to evaluate the risk associated with accepting the decisions of the trading system. At last, the average number of daily operations is used to evaluate the frequency of the entry/exit operations. IV. RESULTS A. Performance measurement The performance of the trained networks is measured by using a pattern set which is distinct from the pattern sets used for training (out-of-sample).this leads to an estimation what would have happened if the trained networks had been used in production, which implies unknown and new data points. B. Stock Market Data: The sample data are used from National Stock Exchange (NSE) during the period from April 01 to Mar 10. We have taken the Ranbaxy Ltd for our Organized by Department Of ECE, ANNA UNIVERSITY COIMBATORE. 357

5 implementation and analysis. Table 1. shows the input model of Ranbaxy Ltd. We should predict the future movement of stock market using the price, time, Moving averages (MA1& MA2) & RSI. Table 1 Input data for Ranbaxy V.CONCLUSION In this paper, we mainly discuss steps and methods of using neural network to predict stock market, including sampling principles, principles of determining the number of node in hidden layers. Then the previous stock market performance with the effective stock data and the fuzzified microeconomic data are processed based on the context-based modeling and vector quantization. Finally, the prediction of the stock market performance with the stock data is defuzzified using the fuzzification model to produce a portfolio performance prediction. The major concern of the study is to develop a system that can predict future prices in the stock markets by taking samples of past prices. The developed system seems to work acceptable. Experiments show that obtained forecasts have about 70% accuracy; this result can be seen as satisfying for such difficult task. According to the prediction trades are simulated, i.e. if the network predicts a positive change this leads to a buy signal; a sell signal is issued when the network predicts a falling price. It was shown in Fig.2. According to these signals the performance increases, if the prediction is correct, if not, the performance decreases. Figure2 indication of the generated sell&buy Signal. References [1][Hie941] Y. Hiemstra. Linear Regression versus Backpropagation to Predict Quarterly Excess Returns In: Pmceedings of the Neuml Networks in the Capital Markets 1994, Pasadena, CA, [2][JK95] J. Ledermann, R. A. Klein. Virtual %ding. Probus Publishing, [3][Ras97] M. Rut. Application of Fuzzy Neural Networks on Financial Problems In: Proceedings of the NAFIpS 97, Syracuse, NY, [4]B.Vanstone and G. Finnie, "An Empirical Methodology for Developing Stockmarket Trading Systems using Artificial Neural Networks,"Expert Systems with Applications, vol. In Press, [5]C.M. Huang, Q. Bi, G.S. Stiles, and R. W. Harris. Fast full search equivalent encoding algorithms for image compression using quantization. IEEE Transactions [6]Pamela C.Cosman, Eve A. Riskin, Rbert M. Gray, Combining Vector Quantization and Histogram Equalization, BULB Journals, Information Processing & Management. Vol. 28 Number 6, 1992 pp [7]G. S. Atsalakis and K. P.Valavanis, "Surveying stock market forecasting techniques - part ii: Soft computing methods," Expert Systems with Applications, vol. In Press, Corrected Proof, Organized by Department Of ECE, ANNA UNIVERSITY COIMBATORE. 358

6 [8] A. P. N. Refenes, A. N. Burgess, and Y. Bentz, "Neural networks in financial engineering: A study in methodology," IEEE Transactions on Neural Networks, vol. 8(6), pp , [9] R. Lawrence, "Using Neural Networks to Forecast Stock Market Prices," University of Manitoba, [10] Y. S. A. Mostafa and A. F. Atiya, "Introduction to financial forecasting," Applied Intelligence, vol. 6(3), pp , [11] T. Z. Tan, C. Quek, and G. S. Ng, "Brain inspired genetic complimentary learning for stock market prediction," In IEEE congress on evolutionary computation, vol. 3, pp , Organized by Department Of ECE, ANNA UNIVERSITY COIMBATORE. 359