Advances in Computational Intelligence and Robotics
Random fluctuations occur in the trend of financial time series due to many macroeconomic factors... more Random fluctuations occur in the trend of financial time series due to many macroeconomic factors. Such fluctuations lead to sudden fall after a constant raise or a sudden rise after a constant fall, which are difficult to predict from previous data points. At the fluctuation point, previous data points that are not too close to the target price adversely influence the prediction trend. Far away points may be ignored and close enough virtual data points are explored and incorporated in order to diminish the adverse prediction trend at fluctuations. From the given data points in the training set, virtual data positions (VDP) can be explored and used to enhance the prediction accuracy. This chapter presents some deterministic and stochastic approaches to explore such VDPs. From the given data points in the training set, VDPs are explored and incorporated to the original financial time series to enhance the prediction accuracy of the model. To train and validate the models, ten real st...
LAP LAMBERT Academic Publishing eBooks, Jul 7, 2012
Since last few decades, the wide use of advanced technologies made massive growth in the availabi... more Since last few decades, the wide use of advanced technologies made massive growth in the availability of data from different resources, which lead to the challenge of extracting useful knowledge for specific applications. Classification is one of the most important tasks in mining knowledge from such resources. Though a wide number of traditional techniques are available for classification, but finding a high performance robust classification method, applicable to different real world problem is difficult. This book provides a set of techniques for classification which has been applied to problems from multiple disciplines such as life science, physical science, social science, business, and games. Design of a wide range of classifiers using Polynomial Neural Network and its hybridization with Particle Swam Optimization, Fuzzy logic, Artificial Neural Network has been presented in this book with their scope and limitations. The algorithms and models presented for the large range of classifiers will certainly help the readers in using it conveniently for their application areas.
Advances in intelligent systems and computing, Aug 26, 2014
High dimensionality and small sample size are the intrinsic nature of microarray data, which requ... more High dimensionality and small sample size are the intrinsic nature of microarray data, which require effective computational methods to discover useful knowledge from it. Classification of microarray data is one of the important tasks in this field of work. Representation of the search space with thousands of genes makes this work much complex and difficult to classify efficiently. In this work, three different stages have been adopted to handle the crush of dimensionality and classify the microarray data. At the first stage, statistical measures are used to remove genes that do not contribute for classification. In the second stage, more noisy genes are removed by considering signal-to-noise ratio (SNR). In the third stage, principal component analysis (PCA) method is used to further reduce the dimension. Finally, these reduced datasets are presented to different classification techniques to evaluate their performance. Here, four different classification algorithms are used such as artificial neural network (ANN), naive Bayesian classifier, multiple linear regression (MLR), and k-nearest neighbor (k-NN) to validate the benefits of three-stage dimensionality reduction. The experimental results show that the use of statistical methods, SNR, and PCA improves the overall performance of the classifiers.
alternative approach tononlinear fitting problem Inthis paper wehaveusedaLocal Linear Wavelet[1,2... more alternative approach tononlinear fitting problem Inthis paper wehaveusedaLocal Linear Wavelet[1,2,3,10,11]. Twokeyproblems indesigning the NeuralNetwork(LLWNN)modelforpattern WNN arehowtodetermine WNN architecture and classification. Thedifference ofthenetwork withwhatlearning algorithm canbeeffectively usedfor conventional wavelet neural network (WNN)isthattraining theWNN [4]. Theseproblems arerelated to theconnection weights between thehidden layer and determine anoptimal WNN architecture, toarrange output layer ofconventional WNN arereplaced bya thewindows ofwavelets, andtofind theproper lou t ..t S Optimisationorthogonal ornonorthogonal wavelet basis. Curse of local linear modlel. Particle SwarmOptimisation diesonlt isa manl unovdpolmi (PSO)technique usedfortraining theLLWNN. dmensionality isa mamly unsolved problem in Simulation results fortheclassification ofdifferent WNN theory, which brings somedifficulties in benchmark datasets showthefeasibility and applying aWNN tohigh dimension problems. effectiveness oftheproposedmethod. Thebasis function neural networks areaclass of neural networks, inwhichtheoutput ofthenetwork
Prediction of stock index remains a challenging task of the financial time series prediction proc... more Prediction of stock index remains a challenging task of the financial time series prediction process. Though different non-linear prediction models are in use, their prediction accuracy does not improve beyond certain level. In order to improve the forecasting accuracy, this paper proposes a chemical reaction optimization based virtual data position (CROVDP) exploration method for financial time series prediction. From the given data points in the training set, CROVDPs are explored by the proposed method and incorporated to the original financial time series to enhance the prediction accuracy of a multilayer perceptron (MLP) based forecasting model. The effectiveness of the proposed method is compared with three deterministic methods such as linear, Lagrange and Taylor interpolation as well as two stochastic methods such as uniform and Gaussian method. To train and validate the models in presence of virtual data positions (VDP), ten real stock indices are used. Simulation results show that the models developed with the proposed CROVDP exploration method yields much better prediction accuracy than others.
2012 International Conference on Computing, Communication and Applications
Forecasting stock exchange rates is an important financial problem that is receiving increasing a... more Forecasting stock exchange rates is an important financial problem that is receiving increasing attention. During the last few years, a number of neural network models and hybrid models have been proposed for obtaining accurate prediction results, in an attempt to outperform the traditional linear and nonlinear approaches. This paper evaluates the effectiveness of neural network models which are known to
2012 International Conference on Computing, Communication and Applications, 2012
Many researchers have suggested different approach to map the thinking about thinking. However li... more Many researchers have suggested different approach to map the thinking about thinking. However little has been done to evaluate it computationally. In this paper, we propose a technique based on Swarm Intelligence to estimate the human foreseeability. The statistics of moral reasoner of UCI machine learning repository has been used to validate this work. A number of rule sets have been developed to assess the human foreseeability to a reasonable extent. Further an analysis is presented on the impact of human features on foreseeability.
International Journal of Industrial and Systems Engineering, 2017
Variations occur in the trends of financial time series data due to several reasons. Such random ... more Variations occur in the trends of financial time series data due to several reasons. Such random fluctuations lead to a sudden fall after a steady increase or a sudden rise after a gradual fall in the trend of financial time series data; this makes it difficult to predict. This research work explores the impact of virtual data positions (VDPs) on financial time series forecasting using three different data exploration techniques with artificial neural network (ANN) trained by back propagation. To train and validate these models, the daily closing prices of BSE, DJIA, NASDAQ, FTSE100, S&P500, TAIEX and SSE indices were used. Since none of the exploration techniques were found to be noticeably superior compared with the others, additional experiments were conducted using genetic algorithm (GA) to train ANN; this resulted in better accuracy.
2021 19th OITS International Conference on Information Technology (OCIT), 2021
Development of a new and robust classification method is yet a hot area of research. The polynomi... more Development of a new and robust classification method is yet a hot area of research. The polynomial neural network(PNN) is one such method. However, its model complexity and high computation time restrict its usage. This paper proposes an evolutionary teaching-learning base modified polynomial classifier(ETL-PC), a modified PNN model which initially starts with a polynomial neural network and then uses an artificial neural network to find the output. The model uses a binary encoded genetic algorithm to select a subset of features and partial descriptions(PDs) relevant for classification and uses teaching-learning based optimization method to determine the appropriate weights. The performance of ETL-PC is compared with the performance of few other classifiers using different metrics and it is observed that ETL-PC performs either better or as other methods in many cases and may be considered as a competitive model.
Handbook of Research on Modeling, Analysis, and Application of Nature-Inspired Metaheuristic Algorithms
This chapter will discuss and present a new model to perform a competency map in an educational i... more This chapter will discuss and present a new model to perform a competency map in an educational institute which has moderate number of faculty members on whom the map has to be performed. Performing such a map is a tough but essential task. Hence utmost objectivity must be followed for the procedure. In this chapter the authors performed academic load assignment to the faculty members of a particular dept at the onset of a semester. Few parameters that have been considered as the input parameters are depth of knowledge, sincerity, class management, contribution towards research, text book publication. part from that one of the main concerns is the assignment is done by taking into consideration the preferences of a particular faculty in terms of teaching a subject. There are number of constraints which need to be considered while making the load assignment. The AICTE guidelines for teaching load allotment have been considered as a baseline. The MOPSO has been used to perform the competency map and the simulation results have been presented to show the effectiveness of the method.
Manual estimation of compressive strength of concrete (CSC) is a time consuming and difficult tas... more Manual estimation of compressive strength of concrete (CSC) is a time consuming and difficult task. It has been an active area of research in the domain of manufacturing engineering. Soft computing techniques are found better to statistical methods applied for prediction of CSC. However, sophisticated prediction models are still missing and need to be explored. Random vector functional link network (RVFLN) significantly reduces the time complexity by assigning input layer weights and bias randomly without further modification. Only, output layer weights are calculated iteratively by gradient methods or non-iteratively by closed form solution like least square methods. It is an efficient algorithm with low time complexity and is able to handle complex domain problems without compromising on accuracy. Motivated from characteristics of RVFLN, in this work we develop a RVFLNbased forecast for estimating compressive strength of concrete cement. A publicly available dataset from UCI repository is used to develop and access the performance of the model. For comparative analysis, few other models such as FLANN, MLP, RBFNN, MLR, and ARIMA are also developed and used for the forecasting considering samples with curing ages at 14, 28, 56, and 91 days. All the models are evaluated in terms of MAPE, ARV, U of Theil's statistics (UT), NMSE, and execution time. Outcomes of the comparative studies and statistical significance tests are in favor of RVFLN-based forecasting.
2012 World Congress on Information and Communication Technologies
With the rise of artificial intelligence technology and the growing interrelated markets of the l... more With the rise of artificial intelligence technology and the growing interrelated markets of the last two decades offering unprecedented trading opportunities, technical analysis simply based on forecasting models is no longer enough. To meet the trading challenge in today's global market, technical analysis must be redefined. Before using the neural network models some issues such as data preprocessing, network architecture and learning parameters are to be considered. Data normalization is a fundamental data preprocessing step for learning from data before feeding to the Artificial Neural Network (ANN). Finding an appropriate method to normalize time series data is not a simple task. This work evaluates various normalization methods used in ANN model trained with gradient descent (ANN-GD), genetic algorithm (ANN-GA), and functional link artificial neural network model trained with GD (FLANN-GD) and genetic algorithm (FLANN-GA). The study is applied on daily closing price of Bombay stock exchange (BSE) and experimental result.
Uncertainty and complexity associated with the stock data make the exact determination of future ... more Uncertainty and complexity associated with the stock data make the exact determination of future prices impossible. Successful prediction of a stock future price requires an efficient prediction system. This paper proposes an artificial chemical reaction optimization based functional link network termed as ACFLN for stock market forecasting. The efficiency of the proposed model has been evaluated by forecasting five real stock market prices such as BSE, DJIA, NASDAQ, TAIEX and FTSE. Different experiments are conducted to evaluate the performance of the proposed model such as forecasting the stock price 1 day ahead, 1 week ahead, and 1 month ahead. Data is obtained for all the working days in a year and for each data the said experiments are conducted. From simulation studies, it is revealed that the proposed model achieves better forecasting accuracies over others.
Advances in Computational Intelligence and Robotics
Random fluctuations occur in the trend of financial time series due to many macroeconomic factors... more Random fluctuations occur in the trend of financial time series due to many macroeconomic factors. Such fluctuations lead to sudden fall after a constant raise or a sudden rise after a constant fall, which are difficult to predict from previous data points. At the fluctuation point, previous data points that are not too close to the target price adversely influence the prediction trend. Far away points may be ignored and close enough virtual data points are explored and incorporated in order to diminish the adverse prediction trend at fluctuations. From the given data points in the training set, virtual data positions (VDP) can be explored and used to enhance the prediction accuracy. This chapter presents some deterministic and stochastic approaches to explore such VDPs. From the given data points in the training set, VDPs are explored and incorporated to the original financial time series to enhance the prediction accuracy of the model. To train and validate the models, ten real st...
LAP LAMBERT Academic Publishing eBooks, Jul 7, 2012
Since last few decades, the wide use of advanced technologies made massive growth in the availabi... more Since last few decades, the wide use of advanced technologies made massive growth in the availability of data from different resources, which lead to the challenge of extracting useful knowledge for specific applications. Classification is one of the most important tasks in mining knowledge from such resources. Though a wide number of traditional techniques are available for classification, but finding a high performance robust classification method, applicable to different real world problem is difficult. This book provides a set of techniques for classification which has been applied to problems from multiple disciplines such as life science, physical science, social science, business, and games. Design of a wide range of classifiers using Polynomial Neural Network and its hybridization with Particle Swam Optimization, Fuzzy logic, Artificial Neural Network has been presented in this book with their scope and limitations. The algorithms and models presented for the large range of classifiers will certainly help the readers in using it conveniently for their application areas.
Advances in intelligent systems and computing, Aug 26, 2014
High dimensionality and small sample size are the intrinsic nature of microarray data, which requ... more High dimensionality and small sample size are the intrinsic nature of microarray data, which require effective computational methods to discover useful knowledge from it. Classification of microarray data is one of the important tasks in this field of work. Representation of the search space with thousands of genes makes this work much complex and difficult to classify efficiently. In this work, three different stages have been adopted to handle the crush of dimensionality and classify the microarray data. At the first stage, statistical measures are used to remove genes that do not contribute for classification. In the second stage, more noisy genes are removed by considering signal-to-noise ratio (SNR). In the third stage, principal component analysis (PCA) method is used to further reduce the dimension. Finally, these reduced datasets are presented to different classification techniques to evaluate their performance. Here, four different classification algorithms are used such as artificial neural network (ANN), naive Bayesian classifier, multiple linear regression (MLR), and k-nearest neighbor (k-NN) to validate the benefits of three-stage dimensionality reduction. The experimental results show that the use of statistical methods, SNR, and PCA improves the overall performance of the classifiers.
alternative approach tononlinear fitting problem Inthis paper wehaveusedaLocal Linear Wavelet[1,2... more alternative approach tononlinear fitting problem Inthis paper wehaveusedaLocal Linear Wavelet[1,2,3,10,11]. Twokeyproblems indesigning the NeuralNetwork(LLWNN)modelforpattern WNN arehowtodetermine WNN architecture and classification. Thedifference ofthenetwork withwhatlearning algorithm canbeeffectively usedfor conventional wavelet neural network (WNN)isthattraining theWNN [4]. Theseproblems arerelated to theconnection weights between thehidden layer and determine anoptimal WNN architecture, toarrange output layer ofconventional WNN arereplaced bya thewindows ofwavelets, andtofind theproper lou t ..t S Optimisationorthogonal ornonorthogonal wavelet basis. Curse of local linear modlel. Particle SwarmOptimisation diesonlt isa manl unovdpolmi (PSO)technique usedfortraining theLLWNN. dmensionality isa mamly unsolved problem in Simulation results fortheclassification ofdifferent WNN theory, which brings somedifficulties in benchmark datasets showthefeasibility and applying aWNN tohigh dimension problems. effectiveness oftheproposedmethod. Thebasis function neural networks areaclass of neural networks, inwhichtheoutput ofthenetwork
Prediction of stock index remains a challenging task of the financial time series prediction proc... more Prediction of stock index remains a challenging task of the financial time series prediction process. Though different non-linear prediction models are in use, their prediction accuracy does not improve beyond certain level. In order to improve the forecasting accuracy, this paper proposes a chemical reaction optimization based virtual data position (CROVDP) exploration method for financial time series prediction. From the given data points in the training set, CROVDPs are explored by the proposed method and incorporated to the original financial time series to enhance the prediction accuracy of a multilayer perceptron (MLP) based forecasting model. The effectiveness of the proposed method is compared with three deterministic methods such as linear, Lagrange and Taylor interpolation as well as two stochastic methods such as uniform and Gaussian method. To train and validate the models in presence of virtual data positions (VDP), ten real stock indices are used. Simulation results show that the models developed with the proposed CROVDP exploration method yields much better prediction accuracy than others.
2012 International Conference on Computing, Communication and Applications
Forecasting stock exchange rates is an important financial problem that is receiving increasing a... more Forecasting stock exchange rates is an important financial problem that is receiving increasing attention. During the last few years, a number of neural network models and hybrid models have been proposed for obtaining accurate prediction results, in an attempt to outperform the traditional linear and nonlinear approaches. This paper evaluates the effectiveness of neural network models which are known to
2012 International Conference on Computing, Communication and Applications, 2012
Many researchers have suggested different approach to map the thinking about thinking. However li... more Many researchers have suggested different approach to map the thinking about thinking. However little has been done to evaluate it computationally. In this paper, we propose a technique based on Swarm Intelligence to estimate the human foreseeability. The statistics of moral reasoner of UCI machine learning repository has been used to validate this work. A number of rule sets have been developed to assess the human foreseeability to a reasonable extent. Further an analysis is presented on the impact of human features on foreseeability.
International Journal of Industrial and Systems Engineering, 2017
Variations occur in the trends of financial time series data due to several reasons. Such random ... more Variations occur in the trends of financial time series data due to several reasons. Such random fluctuations lead to a sudden fall after a steady increase or a sudden rise after a gradual fall in the trend of financial time series data; this makes it difficult to predict. This research work explores the impact of virtual data positions (VDPs) on financial time series forecasting using three different data exploration techniques with artificial neural network (ANN) trained by back propagation. To train and validate these models, the daily closing prices of BSE, DJIA, NASDAQ, FTSE100, S&P500, TAIEX and SSE indices were used. Since none of the exploration techniques were found to be noticeably superior compared with the others, additional experiments were conducted using genetic algorithm (GA) to train ANN; this resulted in better accuracy.
2021 19th OITS International Conference on Information Technology (OCIT), 2021
Development of a new and robust classification method is yet a hot area of research. The polynomi... more Development of a new and robust classification method is yet a hot area of research. The polynomial neural network(PNN) is one such method. However, its model complexity and high computation time restrict its usage. This paper proposes an evolutionary teaching-learning base modified polynomial classifier(ETL-PC), a modified PNN model which initially starts with a polynomial neural network and then uses an artificial neural network to find the output. The model uses a binary encoded genetic algorithm to select a subset of features and partial descriptions(PDs) relevant for classification and uses teaching-learning based optimization method to determine the appropriate weights. The performance of ETL-PC is compared with the performance of few other classifiers using different metrics and it is observed that ETL-PC performs either better or as other methods in many cases and may be considered as a competitive model.
Handbook of Research on Modeling, Analysis, and Application of Nature-Inspired Metaheuristic Algorithms
This chapter will discuss and present a new model to perform a competency map in an educational i... more This chapter will discuss and present a new model to perform a competency map in an educational institute which has moderate number of faculty members on whom the map has to be performed. Performing such a map is a tough but essential task. Hence utmost objectivity must be followed for the procedure. In this chapter the authors performed academic load assignment to the faculty members of a particular dept at the onset of a semester. Few parameters that have been considered as the input parameters are depth of knowledge, sincerity, class management, contribution towards research, text book publication. part from that one of the main concerns is the assignment is done by taking into consideration the preferences of a particular faculty in terms of teaching a subject. There are number of constraints which need to be considered while making the load assignment. The AICTE guidelines for teaching load allotment have been considered as a baseline. The MOPSO has been used to perform the competency map and the simulation results have been presented to show the effectiveness of the method.
Manual estimation of compressive strength of concrete (CSC) is a time consuming and difficult tas... more Manual estimation of compressive strength of concrete (CSC) is a time consuming and difficult task. It has been an active area of research in the domain of manufacturing engineering. Soft computing techniques are found better to statistical methods applied for prediction of CSC. However, sophisticated prediction models are still missing and need to be explored. Random vector functional link network (RVFLN) significantly reduces the time complexity by assigning input layer weights and bias randomly without further modification. Only, output layer weights are calculated iteratively by gradient methods or non-iteratively by closed form solution like least square methods. It is an efficient algorithm with low time complexity and is able to handle complex domain problems without compromising on accuracy. Motivated from characteristics of RVFLN, in this work we develop a RVFLNbased forecast for estimating compressive strength of concrete cement. A publicly available dataset from UCI repository is used to develop and access the performance of the model. For comparative analysis, few other models such as FLANN, MLP, RBFNN, MLR, and ARIMA are also developed and used for the forecasting considering samples with curing ages at 14, 28, 56, and 91 days. All the models are evaluated in terms of MAPE, ARV, U of Theil's statistics (UT), NMSE, and execution time. Outcomes of the comparative studies and statistical significance tests are in favor of RVFLN-based forecasting.
2012 World Congress on Information and Communication Technologies
With the rise of artificial intelligence technology and the growing interrelated markets of the l... more With the rise of artificial intelligence technology and the growing interrelated markets of the last two decades offering unprecedented trading opportunities, technical analysis simply based on forecasting models is no longer enough. To meet the trading challenge in today's global market, technical analysis must be redefined. Before using the neural network models some issues such as data preprocessing, network architecture and learning parameters are to be considered. Data normalization is a fundamental data preprocessing step for learning from data before feeding to the Artificial Neural Network (ANN). Finding an appropriate method to normalize time series data is not a simple task. This work evaluates various normalization methods used in ANN model trained with gradient descent (ANN-GD), genetic algorithm (ANN-GA), and functional link artificial neural network model trained with GD (FLANN-GD) and genetic algorithm (FLANN-GA). The study is applied on daily closing price of Bombay stock exchange (BSE) and experimental result.
Uncertainty and complexity associated with the stock data make the exact determination of future ... more Uncertainty and complexity associated with the stock data make the exact determination of future prices impossible. Successful prediction of a stock future price requires an efficient prediction system. This paper proposes an artificial chemical reaction optimization based functional link network termed as ACFLN for stock market forecasting. The efficiency of the proposed model has been evaluated by forecasting five real stock market prices such as BSE, DJIA, NASDAQ, TAIEX and FTSE. Different experiments are conducted to evaluate the performance of the proposed model such as forecasting the stock price 1 day ahead, 1 week ahead, and 1 month ahead. Data is obtained for all the working days in a year and for each data the said experiments are conducted. From simulation studies, it is revealed that the proposed model achieves better forecasting accuracies over others.
Uploads
Papers by bijan misra