CONSTRAINED FORMULATIONS AND ALGORITHMS FOR PREDICTING STOCK PRICES BY RECURRENT FIR NEURAL NETWORKS

2006 ◽  
Vol 05 (04) ◽  
pp. 639-658 ◽  
Author(s):  
BENJAMIN W. WAH ◽  
MING-LUN QIAN
Keyword(s):  
Time Series ◽  
Constrained Optimization ◽  
Stock Prices ◽  
Finite Impulse Response ◽  
Saddle Points ◽  
Stationary Time Series ◽  
Good Prediction ◽  
Learning Problem ◽  
Good Prediction Accuracy ◽  
Low Pass

In this paper, we develop a new constrained artificial-neural-network (ANN) formulation and the associated learning algorithm for predicting stock prices, a difficult time-series prediction problem. We characterize daily stock prices as a noisy non-stationary time series and identify its predictable low-frequency components. Using a recurrent finite-impulse-response ANN, we formulate the learning problem as a constrained optimization problem, develop constraints for incorporating cross validations, and solve the learning problem using algorithms based on the theory of extended saddle points for nonlinear constrained optimization. Finally, we illustrate our prediction results on ten stock-price time series. Our main contributions in this paper are the channel-specific low-pass filtering of noisy time series obtained by wavelet decomposition, the transformation of the low-pass signals to improve their stationarity, and the incorporation of constraints on cross validation that can improve the accuracy of predictions. Our experimental results demonstrate good prediction accuracy and annual returns.

scholarly journals Low-Pass Filtering Method for Poisson Data Time Series

2021 ◽  
Vol 11 (10) ◽  
pp. 4524
Author(s):  
Victor Getmanov ◽  
Vladislav Chinkin ◽  
Roman Sidorov ◽  
Alexei Gvishiani ◽  
Mikhail Dobrovolsky ◽  
...  
Keyword(s):  
Time Series ◽  
Finite Impulse Response ◽  
Digital Processing ◽  
Distributed Data ◽  
Gaussian Filter ◽  
Experimental Physics ◽  
Filtering Method ◽  
Filter Synthesis ◽  
Low Pass ◽  
Poisson Data

Problems of digital processing of Poisson-distributed data time series from various counters of radiation particles, photons, slow neutrons etc. are relevant for experimental physics and measuring technology. A low-pass filtering method for normalized Poisson-distributed data time series is proposed. A digital quasi-Gaussian filter is designed, with a finite impulse response and non-negative weights. The quasi-Gaussian filter synthesis is implemented using the technology of stochastic global minimization and modification of the annealing simulation algorithm. The results of testing the filtering method and the quasi-Gaussian filter on model and experimental normalized Poisson data from the URAGAN muon hodoscope, that have confirmed their effectiveness, are presented.

scholarly journals VISUAL ANALYSIS OF RECURRENCE OF TIME SERIES OF THE COORDINATES ENU IN THE GPS STATIONS

2018 ◽  
Vol 24 (4) ◽  
pp. 470-484
Author(s):  
Alfonso Tierra ◽  
Rubén León ◽  
Alexis Tinoco-S ◽  
Carolina Cañizares ◽  
Marco Amores ◽  
...  
Keyword(s):  
Time Series ◽  
Continuous Monitoring ◽  
Visual Analysis ◽  
Finite Impulse Response ◽  
Satellite System ◽  
Nearest Neighbours ◽  
Low Pass ◽  
Global Navigation Satellite ◽  
Low Pass Filters ◽  
Content Information

Abstract The time series content information about the dynamic behavior of the system under study. This behavior could be complex, irregular and no lineal. For this reason, it is necessary to study new models that can solve this dynamic more satisfactorily. In this work a visual analysis of recurrence from time series of the coordinate’s variation ENU (East, North, Up) will be made. This analysis was obtained from nine continuous monitoring stations GPS (Global Navigation Satellite System); the intention is to study their behavior, they belong to the Equatorian GPS Network that materializes the reference system SIRGAS - ECUADOR. The presence of noise in the observations was reduced using digital low pass filters with Finite Impulse Response (FIR). For these series, the time delay was determined using the average mutual information, and for the minimum embedding dimension the False Nearest Neighbours (FNN) method was used; the purpose is to obtain the recurrent maps of each coordinates. The results of visual analysis show a strong tendency, especially in the East and North coordinates, while the Up coordinates indicate discontinued, symmetric and periodic behavior.

scholarly journals Fast Prediction with Sparse Multikernel LS-SVR Using Multiple Relevant Time Series and Its Application in Avionics System

2015 ◽  
Vol 2015 ◽  
pp. 1-10
Author(s):  
Yang M. Guo ◽  
Pei He ◽  
Xiang T. Wang ◽  
Ya F. Zheng ◽  
Chong Liu ◽  
...  
Keyword(s):  
Time Series ◽  
Prediction Model ◽  
Prediction Accuracy ◽  
Mean Squared Error ◽  
Computation Time ◽  
Support Vector ◽  
Good Prediction ◽  
Trend Prediction ◽  
Good Prediction Accuracy ◽  
Fast Prediction

Health trend prediction is critical to ensure the safe operation of highly reliable systems. However, complex systems often present complex dynamic behaviors and uncertainty, which makes it difficult to develop a precise physical prediction model. Therefore, time series is often used for prediction in this case. In this paper, in order to obtain better prediction accuracy in shorter computation time, we propose a new scheme which utilizes multiple relevant time series to enhance the completeness of the information and adopts a prediction model based on least squares support vector regression (LS-SVR) to perform prediction. In the scheme, we apply two innovative ways to overcome the drawbacks of the reported approaches. One is to remove certain support vectors by measuring the linear correlation to increase sparseness of LS-SVR; the other one is to determine the linear combination weights of multiple kernels by calculating the root mean squared error of each basis kernel. The results of prediction experiments indicate preliminarily that the proposed method is an effective approach for its good prediction accuracy and low computation time, and it is a valuable method in applications.

scholarly journals Dependence structure in financial time series: Applications and evidence from wavelet analysis

2021 ◽  
Author(s):  
◽  
Long Hai Vo
Keyword(s):  
Time Series ◽  
Wavelet Analysis ◽  
Stock Prices ◽  
Long Memory ◽  
Stationary Time Series ◽  
Dependence Structure ◽  
Garch Models ◽  
Parameter Estimates ◽  
Hurst Index ◽  
Low Frequencies

<p>Conventional time series theory and spectral analysis have independently achieved significant popularity in mainstream economics and finance research over long periods. However, the fact remains that each is somewhat lacking if the other is absent. To overcome this problem, a new methodology, wavelet analysis, has been developed to capture all the information localized in time and in frequency, which provides us with an ideal tool to study non-stationary time series. This paper aims to explore the application of a variety of wavelet-based methodologies in conjunction with conventional techniques, such as the Generalized Autoregressive Conditional Heteroskedasticity (GARCH) models and long-memory parameter estimates, in analysing the short and long term dependence structure of financial returns and volatility. Specifically, by studying the long-memory property of these time series we hope to identify the source of their possible predictability. Above all else, we document the indispensable role of trading activities associated with low frequencies in determining the long-run dependence of volatility. It follows that GARCH models incorporating long-memory and asymmetric returns-volatility dynamics can provide reasonably accurate volatility forecasts. Additionally, the persistence parameter of returns, represented by the Hurst index, is observed to be correlated to trading profits obtained from typical technical rules designed to detect and capitalize on existing trending behaviour of stock prices. This implies that the Hurst index can be used as a good indicator of the long-memory characteristic of the market, which in turn drives such trending behaviour.</p>

scholarly journals Dependence structure in financial time series: Applications and evidence from wavelet analysis

2021 ◽  
Author(s):  
◽  
Long Hai Vo
Keyword(s):  
Time Series ◽  
Wavelet Analysis ◽  
Stock Prices ◽  
Long Memory ◽  
Stationary Time Series ◽  
Dependence Structure ◽  
Garch Models ◽  
Parameter Estimates ◽  
Hurst Index ◽  
Low Frequencies

<p>Conventional time series theory and spectral analysis have independently achieved significant popularity in mainstream economics and finance research over long periods. However, the fact remains that each is somewhat lacking if the other is absent. To overcome this problem, a new methodology, wavelet analysis, has been developed to capture all the information localized in time and in frequency, which provides us with an ideal tool to study non-stationary time series. This paper aims to explore the application of a variety of wavelet-based methodologies in conjunction with conventional techniques, such as the Generalized Autoregressive Conditional Heteroskedasticity (GARCH) models and long-memory parameter estimates, in analysing the short and long term dependence structure of financial returns and volatility. Specifically, by studying the long-memory property of these time series we hope to identify the source of their possible predictability. Above all else, we document the indispensable role of trading activities associated with low frequencies in determining the long-run dependence of volatility. It follows that GARCH models incorporating long-memory and asymmetric returns-volatility dynamics can provide reasonably accurate volatility forecasts. Additionally, the persistence parameter of returns, represented by the Hurst index, is observed to be correlated to trading profits obtained from typical technical rules designed to detect and capitalize on existing trending behaviour of stock prices. This implies that the Hurst index can be used as a good indicator of the long-memory characteristic of the market, which in turn drives such trending behaviour.</p>

Sign in / Sign up

Export Citation Format

Share Document