Copy Right to GARPH Page 38

Transcription

1 IMPLEMENT EFFICIENT ALGORITHM FOR PREDICTIONS OF STOCK MARKET BY MULTICLASS SUPPORT VECTOR MACHINE 1 VAIBHAV INGOLE Department of Computer Science & Engineering, TIT, Bhopal, India vingole42@rediffmail.com 2 Dr. BHUPESH GAUR Associate Professor, Department of Computer Science & Engineering, TIT, Bhopal, India 3 PROF. DEEPAK SINGH TOMAR Assistant Professor, Department of Computer Science & Engineering, TIT, Bhopal, India ABSTRACT: The Stock Exchange Share Value is very hard to predict as there are no significant rules to estimate the movement of stock. Stock price prediction is most rising fields of research and many methods like technical analysis, statistical analysis, time series analysis etc are applied for this purpose. Artificial Neural Network is a popular technique for the stock price prediction. Support Vector Machine (SVM) is a comparatively new learning algorithm that has the desirable characteristics of control of the decision functions, the use of the kernel method, and the sparsity of the solution. In this paper, presenting a theoretical and empirical framework to apply the Support Vector Machines strategy to predict the stock market. SVM work on large data set value also by comparing support vector machine predicted output with artificial neural network predicted output. Support Vector Machine is used for analyzing the relationship of those factors which predicting the stock market performance. Our results suggest that SVM is a powerful predictive means for stock predictions in the financial marketplace. Key words: stock classification, data mining, support vector machine, forecasting, multivariate classification. 1. INTRODUCTION The macro-economic environment and the financial markets are complex, evolutionary, and non-linear dynamical systems. Before going to study the historic volatile days over the decade, let us first know what are: a) Stock Markets b) Stock exchanges a) Stock Markets: The Stock Market is a marketplace where the trade of company stock, both listed securities and unlisted takes place. It is unlike from stock exchange because it includes all the national stock exchanges of a country. For example, we exercise the name, "the stock market was up today" or "the stock market bubble. b) Stock Exchanges: The Stock Exchanges is a structured marketplace, either corporation or mutual organization. Organization members gather to buy and sell company stocks or additional securities. The members may work either as agents for their customers, or as principals for their individual accounts. Stock exchanges also help for issuing and redemption of securities and other financial instrument including the compensation of income and dividend. The record keeping is central but trade is linked to such physical place since modern market is computerized. The trade on an exchange is only done by members and stock broker who have a seat on the exchange [15]. 2. LITERATURE SURVEY Zen Hu et. al. introduced the prediction of stock market with the help of Support vector Machine. These three scantiest is working together for just beginning the first proposal for which we predict the stock market value with the help of support vector machine. It is a relatively new learning algorithm that has the desirable properties of the control of the decision function, the use of the kernel methods, and the sparcity of the solutions. They present a theoretical and empirical framework for application the Support Vector Machines strategy to predict the stock market. First, four company-specific and six macroeconomic factor that may affect the stock trend are selected for furthermore stock multivariate analysis. And secondly, Support Vector Machine is used in analyzing the relationship of these factors and forecasting the stock performance. They giving outcomes propose that SVM is a powerful predictive tool for stock predictions in the financial area [1]. C.J.C. Burges. He describe the SVM for separable and non separable data, working through the non trivial example details. He describes a mechanical analogy and discuss Copy Right to GARPH Page 38

2 when SVM solution are unique and when they are global. Also he describe the how support vector training can be sensibly implemented and giving details the kernel map procedure which is used to construct SVM Solution which are in non linear data. They proves that SVM have very large radial basis function kernel [3]. Cortes et. al. they works on the non separable training data because the prior work is done on the training dataset which is separated by error. They also compare the support vector network to rest of the learning algorithm [4]. Alex J. Smola et. al. overview of the basic thoughts underlying Support Vector (SV) machines for function evaluation. Furthermore, They comprise a summary of presently used algorithms or training SV machines, casing both the quadratic programming part and advanced methods for concerning with large no of datasets. Finally, they mention some modification and extensions that have been applied to the standard SV algorithm, and argue the aspect of regularization from a SV perspective [10]. Tay and Lijuan observed the possibility of SVM in financial time series forecasting by comparing it with a multi-layer back-propagation (BP) neural network. Investigation of the experimental results proved that it is advantageous the application of SVMs to forecast financial time series. Fan and Palaniswami, (2001) used SVMs for classification for stock selection on the Australian Stock Exchange and significantly outperformed the benchmark [12]. 3. PROBLEM DEFINITION For predicting stock market value the existing work was done by using Artificial Neural Network (ANN) which work with large data set but having predicted result is not efficiently also there is the problem of over fitting. This system relative very complex and each time of iteration we have to train the neurons. It gives less time but predicted output is not good, it produces efficiency error. 4. PROPOSED ARCHITECTURE The stock market their various algorithm is working out for the better result we have to build the algorithm to predict the stock market value by using support vector Machine. Basically the SVM is working on the Machine learning algorithm. SVM gets the large amount data and classify that data into various group of sets in the plane. To predict the stock market for any stock their large amount of dataset available globally on the financial website she result. T o to collect data base from the desired order and that dataset classify according to the result generation. Figure:4.1 shows the block diagram of the proposed solution of the project. Figure1: System Architecture The dataset in the form of.csv file that file support the data system get that data and perform the operation on it. There are two dataset we can use in the project first data is the web data that data can download from the financial website and second dataset is generalized data set both the dataset can convert into the.csv, and. xlsx file for the prediction of stock market. Further that data can goes for the preprocessing and then the SVM classifier classify that data and after that it predict the value of the given stock. Here we are buildup the system for the two algorithm SVM and ANN so we compare the dataset in both algorithm. And both algorithms predict the value of stock market. And finally we conclude the result of the prediction. 5. PROPOSED ALGORITHM Stepwise description of algorithm for Prediction of stock market by using SVM:- 1. Start 2. Select the stock. 3. Download the stock value up to the current Date. 4. Convert the download data set in the form of.csv or.xlsx form. 5. System takes 70% data set for the Training Phase and remaining 30% data set for Training Phase 6. Apply dataset to the SVM algorithm. 7. Get graph before mapping. 8. Get graph after mapping 9. Graph generate with minimize mean square value. 10. Graph generate with less number of error. 11. Finally, generate the graph with original value along with Predicted value. 12. Generate elapse time for executing algorithm. 13. Exit Copy Right to GARPH Page 39

3 6. RESULT ANALYSIS Figure 4: Stock values after mapping Figure 2: Selecting dataset file from the dataset in SVM After click in on the SVM button the algorithm asked for he input dataset. Now select the appropriate path when the dataset is to be stored. In above screenshot window we are selecting path MATLAB/Project Work/umesh/ford_1991_2015 data set file which going to predict the FORD stock value in year from 01/01/1991 to the 20/04/ When we predict the stock value using SVM firstly it can calculate the errors and algorithm are recursively executed for minimizing the error. Actually it can minimize the mean square error to plotting the predicted graph with the help of lo 2 c and log 2 g function. Screenshots shown the result for minimize the mean square error in given stock value. Figure 5: Minimize the Mean Square Error using lo 2 C and log 2 g graph Figure 3: Stock value before mapping After selecting stock value it will shows the graph for the stock value before mapping as shown in screenshot. And the next window shows the graph that value after mapping. Finally the algorithm predicts the stock market value foe the stock Ford in year from 1991 to The graph shows the original value as well as the predicted value for the stock. The red line shows the predicted value and blue line shows the original value. It graph shows very efficient result. Copy Right to GARPH Page 40

4 Figure 6: Generate the testing value with minimum error. Figure 8: Selecting dataset file from the dataset in ANN Finally the algorithm predict the stock market value foe the stock Ford in year from 1991 to The graph shows the original value as well as the predicted value for the stock. The red line shows the predicted value and blue line shows the original value. It graph shows very efficient result, Result efficiency for SVM SVM Algorithm Efficiency Rate for Testing Phase = 100 x 91.13% = Figure7: Prediction of stock Market using SVM Result for prediction stock value using ANN Result efficiency for ANN ANN Algorithm Efficiency Rate for Testing Phase = 100 x When the project is run to forecast the stock value using ANN it shows the a winwdo as shown in screen shot. It mention all the term. here a button is shown if we click on that button. After click on the ANN button the algorithm asked for he input datasets. Now to select the suitable path when the dataset is to be stored. In above screenshot window we are selecting path MATLAB/Project Work/umesh/ford_1991_2015 data set file which going to predict the FORD stock value in year from 01/01/1991 to the 20/04/ % 7. CONCLUSION & FUTURE SCOPE 7.1 CONCLUSIONS In the project, we proposed the use of data collected from various global financial markets with machine learning algorithms to predict the stock index movements. Our conclusion can be summarized into following aspects: SVM algorithm work on the large dataset value which collected from different global financial markets. Also SVM does not give a problem of over fitting. Correlation analysis indicates strong interconnection between the Market stock index and global markets that close right before or at the very beginning of trading time. = Copy Right to GARPH Page 41

5 Various machine learning based models are proposed for predicting daily trend of Market stocks. Numerical results suggests high efficiency A practical trading model is built upon our well trained predictor. The model thus generates higher profit compared to selected benchmarks. In the last we compare the same dataset with SVM and ANN and we concluded that SVM provides the More efficiency than ANN. so SVM predicted result giving is more profitable to the user. [7]N. Ancona, Classification Properties of Support Vector Machines for Regression, Technical Report, RIIESI/CNR- Nr. 02/99. [8]Jianwen Xie, Jianhua Wu, Qingquan Qian: Feature Selection Algorithm Based on Association Rules Mining Method. ACIS-ICIS 2009: [9]T. Joachims, Making Large-Scale SVM Learning Practical, Technical Report, L 8-24, Computer Science Department, University of Dortmund, FUTURE SCOPE In future there is lot of scope to improve the Efficiency of Stock Market Prediction. Lots of research is going on this topic to predict stock value of next day. but SVM is the new advance technology that gives result with very efficient manner. This is the latest topic for research work in prediction of stock market using SVM. Actually there are various application of SVM in different domain. But, in data mining the SVM provides tremendous scope in future. In future SVM provide the predicted output of any stock with efficiently and accurately. Also in future Stock Market prediction app is available in mobile or tab to predict the value of stock in the next day. 8. REFERENCES [1]Zhen Hu, Jie Zhu, and Ken Tse Stocks Market Prediction Using Support Vector Machine th International Conference on Information Management, Innovation Management and Industrial Engineering [2]Wei Huang, Yoshiteru Nakamori, Shou-Yang Wang, Forecasting stock market movement direction with support vector machine, Computers & Operations Research, Volume 32, Issue 10, October 2005, Pages [10]A.J. Smola and B. Scholkopf, A Tutorial on Support Vector Regression, NEUROCOLT2 Technical Report Series, NC2-TR , [11]Debashish Das and Mohammad shorif uddin data mining and neural network techniques in stock market prediction: a methodological review, nternational journal of artificial intelligence & applications (ijaia), vol.4, no.1, january 2013 [12]Mehpare Timor, Hasan Dincer and Şenol Emir doing a work about the Performance comparison of artificial neural network (ANN) and support vector machines (SVM) models for the stock selection problem: An application on the Istanbul Stock Exchange (ISE) African Journal of Business Management Vol. 6(3), pp , 25 January, 2012 [13]Atanu Pal, Diptarka Chakraborty Prediction of Stock Exchange Share Price using ANN and PSO in International journal [14]Sudarsan Padhy and Shom Prasad Das Support Vector Machine for prediction of future nternational Journal of Computer Applications ( ) Volume 41 No.3, March [3]C.J.C. Burges, A Tutorial on Support Vector Machines for Pattern Recognition, Data Mining and Knowledge Discovery, Volume 2, pp. 1-43, Kluwer Academic Publishers, Boston, [4]C. Cortes and V. Vapnik, Support Vector Networks, Machine Learning, 20, , [5]M.Pontil and A. Verri, Properties of Support Vector Machines, Technical Report, Massachusetts Institute of Technology, [6]E.E. Osuna, R. Freund and F. Girosi, Support Vector Machines: Training and Applications, Technical Report, Massachusetts Institute of Technology, Artificial Intelligence Laboratory, AI Memo No. 1602,, Copy Right to GARPH Page 42