Rajashree Dash

A self adaptive differential harmony search based optimized extreme learning machine for financial time series prediction

Abstract This paper proposes a hybrid learning framework called Self Adaptive Differential Harmony Search Based Optimized Extreme Learning Machine (SADHS-OELM) for single hidden layer feed forward neural network (SLFN). The new learning paradigm seeks to take advantage of the generalization ability of extreme learning machines (ELM) along with the global learning capability of a self adaptive differential harmony search technique in order to optimize the fitting performance of SLFNs. SADHS is a variant of harmony search technique that uses the current to best mutation scheme of DE in the pitch adjustment operation for harmony improvisation process. SADHS has been used for optimal selection of the hidden layer parameters, the bias of neurons of the hidden-layer, and the regularization factor of robust least squares, whereas ELM has been applied to obtain the output weights analytically using a robust least squares solution. The proposed learning algorithm is applied on two SLFNs i.e. RBF and a low complexity Functional link Artificial Neural Networks (CEFLANN) for prediction of closing price and volatility of five different stock indices. The proposed learning scheme is also compared with other learning schemes like ELM, DE-OELM, DE, SADHS and two other variants of harmony search algorithm. Performance comparison of CEFLANN and RBF with different learning schemes clearly reveals that CEFLANN model trained with SADHS-OELM outperforms other learning methods and also the RBF model for both stock index and volatility prediction.

Efficient stock price prediction using a Self Evolving Recurrent Neuro-Fuzzy Inference System optimized through a Modified technique

A new neuro-fuzzy network for financial time series prediction is presented.A modified Differential Harmony Search algorithm is used for weight updating.Local as well as delayed output feedback are used for more accurate forecast.Superior predictive ability test is also used for the proposed SERNFIS model. This paper proposes a new Self Evolving Recurrent Neuro-Fuzzy Inference System (SERNFIS) for efficient prediction of highly fluctuating and irregular financial time series data like stock market indices over varying time frames. The network is modeled including the first order Takagi Sugeno Kang (TSK) type fuzzy if then rules with two types of feedback loops. The recurrent structure in the proposed model comes from locally feeding the firing strength of the fuzzy rule back to itself and by including a few time delay components at the output layer. The novelty of the model is based on the fact that the internal temporal feedback loops and time delayed output feedback loops are used for further enhancing the prediction capability of traditional neuro-fuzzy system in handling more dynamic financial time series data. Another recurrent functional link artificial neural network (RCEFLANN) model is also presented for a comparative study. In the second part of the paper a modified differential harmony search (MDHS) technique is proposed for estimating the parameters of the model including the antecedent, consequent and feedback loop parameters. Experimental results obtained by implementing the model on two different stock market indices demonstrate the effectiveness of the proposed model compared to existing models for stock price prediction.

A differential harmony search based hybrid interval type2 fuzzy EGARCH model for stock market volatility prediction

In this paper a new hybrid model integrating an interval type2 fuzzy logic system (IT2FLS) with a computationally efficient functional link artificial neural network (CEFLANN) and an Exponential Generalized Autoregressive Conditional Heteroskedasticity (EGARCH) model has been proposed for accurate forecasting and modeling of financial data with changing variance over time. The proposed model denoted as IT2F-CE-EGARCH helps to enhance the ability of EGARCH model through a joint estimation of the important features of EGARCH like leverage effect, asymmetric shock by leverage effect with the secondary membership functions of interval type2 TSK FLS and the functional expansion and learning component of a CEFLANN. The secondary membership functions with upper and lower limits of IT2FLS provide a forecasting interval for handling more complicated uncertainties involved in volatility forecasting compared to type1 FLS. The performance of the proposed model has been observed with two membership functions i.e. Gaussian with fixed mean, uncertain variance and Gaussian with fixed variance and uncertain mean. The proposed model has also been compared with a few other fuzzy time series models and GARCH family models based on four performance metrics: MSFE, RMSFE, MAFE and Rel MAE. Again a differential harmony search (DHS) algorithm has been suggested for optimizing the parameters of all the fuzzy time series models. The results indicate that the proposed IT2F-CE-EGARCH model offers significant improvements in volatility forecasting performance in comparison with all other specified models over BSE Sensex and CNX Nifty dataset. A hybrid interval type2 Fuzzy-FLANN EGARCH model is proposed to forecast the volatility of three stock market indexes.The TSK-type interval type2 fuzzy inference system uses FLANN in the consequent part of the fuzzy rules for improved mapping.The leverage effect, asymmetric shock by leverage effect of EGARCH model are important for forecasting.A harmony search (HS) based learning strategy is used for EGARCH-FLANN model parameters.Statistical tests are also included for choosing the right model.

Prediction of Financial Time Series Data using Hybrid Evolutionary Legendre Neural Network: Evolutionary LENN

In this paper a predictor model using Legendre Neural Network is proposed for one day ahead prediction of financial time series data. The Legendre Neural Network LENN is a single layer structure that possess faster convergence rate and reduced computational complexity by increasing the dimensionality of the input pattern with a set of linearly independent nonlinear functions. The parameters of the LENN model are estimated using a Moderate Random Search Particle Swarm Optimization Method HMRPSO. The HMRPSO is a variant of PSO that uses a moderate random search method to enhance the global search ability of particles and increases their convergence rates by focusing on valuable search space regions. Training LENN using HMRPSO has also been compared with Particle Swarm Optimization PSO and Differential Evolution DE based learning of LENN for predicting the Bombay Stock Exchange and S&P 500 data sets.

An evolutionary hybrid Fuzzy Computationally Efficient EGARCH model for volatility prediction

Framework of the proposed Fuzzy CE-EGARCH modelDisplay Omitted A hybrid Fuzzy-FLANN EGARCH model is proposed to forecast the volatility of three stock market indexes.The TSK-type fuzzy inference system uses FLANN in the consequent part of the fuzzy rules for improved mapping.The leverage effect, asymmetric shock by leverage effect of EGARCH model are important for forecasting.A differential evolution based learning strategy is used for EGARCH-FLANN model parameters.Multistep prediction and statistical tests are also included. Accurate modeling for forecasting of stock market volatility is a widely interesting research area both in academia as well as financial markets. This paper proposes an innovative Fuzzy Computationally Efficient EGARCH model to forecast the volatility of three stock market indexes. The proposed model represents a joint estimation of the membership function parameters of a TSK-type fuzzy inference system along with the leverage effect, asymmetric shock by leverage effect of EGARCH model in forecasting highly nonlinear and complicated financial time series model more accurately. Further unlike the conventional TSK type fuzzy neural network the proposed model uses a functional link neural network (FLANN) in the consequent part of the fuzzy rules to provide an improved mapping. Moreover, a differential evolution (DE) algorithm is suggested to solve the parameters estimation problem of Fuzzy Computationally Efficient EGARCH model. Being a parallel direct search algorithm, DE has the strength of finding global optimal solutions regardless of the initial values of its few control parameters. Furthermore, the DE based algorithm aims to achieve an optimal solution with a rapid convergence rate. The proposed model has been compared with some GARCH family models and hybrid fuzzy systems and GARCH models based on three performance metrics: MSFE, RMSFE, and MAFE. The results indicate that the proposed method offers significant improvements in volatility forecasting performance in comparison with all other specified models.

A Comparative Study of Chebyshev Functional Link Artificial Neural Network, Multi-layer Perceptron and Decision Tree for Credit Card Fraud Detection

With introduction of online transaction the fraudulent activities through World Wide Web have increased rapidly. Its not only affecting common people but also making them lose huge amount of money. Online transaction basically takes place between merchant and customer, and in this case neither customer nor the card needs to be present at the time of transaction so merchant does not know that whether the customer in the other end is an authorized person or fraudster, so it may lead to an unusual transaction. This kind of online transaction can be easily done using stolen credit card information of a cardholder. To detect status of the current transaction it is imperative to analyze all the previous transactions made by a genuine card holder to know the kind of pattern he/she uses. Based on these patterns new transaction can be categorized as either fraud or legal. There are few data mining techniques which help us to detect a certain pattern on complex and large data sets. In this paper it is proposed to compare Decision Tree, Multi-Layer Perceptron (MLP) and Chebyshev functional link artificial neural network (CFLANN) in terms of their classification accuracy and elapsed time for credit card fraud detection.

Stock price index movement classification using a CEFLANN with extreme learning machine

Recently, an efficient learning algorithm called extreme learning machine (ELM) has been proposed for training of single hidden layer feed forward neural networks (SLFNs). ELM has shown good generalization performances for many real applications with an extremely fast learning speed. This study proposes a computational efficient functional link artificial neural network (CEFLANN) trained with ELM for addressing the problem of classifying stock price index movements as up and down movements. The proposed model is also compared with two other popular networks like Chebyshev FLANN and RBF, used mostly in classification problems. Again the performance of all the networks are compared with back propagation and ELM based learning over two benchmark financial data sets. Experimental results show that the performance of proposed model outperforms the other models.

A comparative study of radial basis function network with different basis functions for stock trend prediction

This paper proposes a radial basis function (RBF) network trained using ridge extreme learning machine to predict the future trend from the past stock index values. Here the task of predicting future stock trend i.e. the up and down movements of stock price index values is cast as a classification problem. Recently extreme learning machine (ELM) is used as an efficient learning algorithm for single hidden layer feed forward neural networks (SLFNs). ELM has shown good generalization performances for many real applications with an extremely fast learning speed. To achieve better performance, an improved ELM with ridge regression called ridge ELM (RELM) is proposed in the study. Gaussian function is the most popular basis function used for RBFN in many applications. But the basis function may not be appropriate for all the applications. Hence the effect of the RBF network with seven different basis functions is compared for addressing the classification task. Again the performance of the RBF network is also compared with back propagation and ELM based learning over two benchmark financial data sets. Experimental results show that evaluating all recognized basis functions suitable for RBF networks is advantageous.

Performance analysis of a higher order neural network with an improved shuffled frog leaping algorithm for currency exchange rate prediction

Abstract Accurate and unbiased prediction of future currency exchange rate is always a potential field of research in domain of financial time series analysis. In this paper, an attempt is urged to examine the predictability of a higher order neural network called Pi-Sigma network for forecasting the highly non linear and dynamic currency exchange rates. An Improved Shuffled Frog Leaping (ISFL) algorithm is set forth to estimate the unrevealed parameters of the network. The network is also examined with few other meta-heuristic learning techniques and compared with other state of art models. Empirically the model validation is realized over three currency exchange data sets such as USD/CAD, USD/CHF, and USD/JPY accumulated within same period of time. Practical analysis of results suggests that the Pi-Sigma network learned with ISFL is a promising predictor model for currency exchange rate prediction compared to other models included in the study.

Chapter 25 – MDHS–LPNN: A Hybrid FOREX Predictor Model Using a Legendre Polynomial Neural Network with a Modified Differential Harmony Search Technique

Abstract This chapter outlines the use of a high order neural network with learning based on a new meta-heuristic optimization algorithm for developing a hybrid FOREX predictor model. The novelty of the work lies in exposing a high order single layer neural network structured using Legendre polynomials for carving an intelligent FOREX predictor model. Further the unknown parameters of the model are estimated using a Modified Differential Harmony Search (MDHS) technique. Modified differential harmony search technique is a new version of original Harmony Search algorithm, in which the current to best mutation strategy is applied in the pitch adjustment operation and instead of using fixed control parameters, they are adapted iteratively according to their previous successful experience. The modified approach leads to an improvement of the convergence speed of the network as well as the predictive ability of the network. Empirically the proposed model is validated by applying it for prediction of currency exchange rates of US Dollar (USD) against four other currencies: Australian Dollar (AUD), British Pound (GBP), Indian Rupee (INR), and Japanese Yen (JPY). From the model verification, it is demonstrated that the proposed network not only provides a higher degree of forecasting accuracy with MDHS learning technique but also performs statistically better than other evaluated learning techniques included in the study.