Financial time series modelling with discounted least squares backpropagation

doi:10.1016/S0925-2312(96)00005-7

Neurocomputing

Volume 14, Issue 2, 5 February 1997, Pages 123-138

https://doi.org/10.1016/S0925-2312(96)00005-7 Get rights and content

Abstract

We propose a simple modification to the error backpropagation procedure which takes into account gradually changing input-output relations. The procedure is based on the principle of Discounted least squares whereby learning is biased towards more recent observations with long term effects experiencing exponential decay through time. This is particularly important in systems in which the structural relationship between input and response vectors changes gradually over time but certain elements of long term memory are still retained. The procedure is implemented by a simple modification of the least-squares cost function commonly used in error backpropagation. We compare the performance of the two cost functions using both a controlled simulation experiment and a non-trivial application in estimating stock returns on the basis of multiple factor exposures. We show that in both cases the DLS procedure gives significantly better results. Typically, there is an average improvement of above 30% (in MSE terms) for the stock return modelling problem.

References (9)

A. Banerjee et al.
W.A. Barnett et al.
P.L. Bartlett
Learning with a slowly changing distribution
C.R. Bonini et al.
Forecasting by smoothed regression

There are more references available in the full text version of this article.

Cited by (36)

Three-level network analysis of the North American natural gas price: A multiscale perspective
2020, International Review of Financial Analysis
Citation Excerpt :
Therefore, it is suitable to study the North American natural gas market from a multiscale perspective. There are many methods for studying financial time series, such as discounted least squares backpropagation (Refenes, Bentz, Bunn, et al., 1997), mixed multiple linear regression (Kalashnikov, Matis, & Pérez-Valdés, 2010), Hilbert-Huang transform (Huang, Wu, Qu, et al., 2003), Diag-BEKK model (Boldanov, Degiannakis, & Filis, 2016) and rough set model (Yao & Herbert, 2009). But these are all focused on the time series itself.
With the US being the largest natural gas producer, the North America formed a mature and leading natural gas market from which emerging markets can learn to develop themselves. It also influences the global market and under various factors, the market presents multiscale characteristics. We explore North American natural gas price, from 1997 to 2018, using wavelet transform and multiscale fluctuation mode network (MFMN) where nodes are a set of fluctuating states extracted from wavelet power and edges are the transmission relationships between the set of states. The multiscale characteristics are analyzed in three levels: macro-network, meso-community and micro-motif. From macroscopic, 131 out of 474 nodes are identified as major contributors of price fluctuation states in the market by weighted out-degree and transmit nodes are identified by betweenness centrality. From mesoscopic, we find that price fluctuation states in the market could be divided into 16 groups and it could be the result of middle- or long-term factors, such as season, weather and policy. From microscopic, 13 directed three-node motifs are detected to present different price fluctuation in continuous short period. And there exist price fluctuation patterns in the market by exploring the community preference for motif containment and motif preference for characters.
Using Volume Weighted Support Vector Machines with walk forward testing and feature selection for the purpose of creating stock trading strategy
2015, Expert Systems with Applications
Citation Excerpt :
In many further studies, an attempt was made to increase models’ performance. Refenes, Bentz, Bunn, Burgess, and Zapranis (1997) proposed a modification of least-squares cost function used in error backpropagation. Obtained estimator was biased towards the more recent observations, and the bias was determined by the decay parameter in sigmoid function.
This study aims to verify whether modified Support Vector Machine classifier can be successfully applied for the purpose of forecasting short-term trends on the stock market. As the input, several technical indicators and statistical measures are selected. In order to conduct appropriate verification dedicated system with the ability to proceed walk-forward testing was designed and developed. In conjunction with modified SVM classifier, we use Fishers method for feature selection. The outcome shows that using the example weighting combined with feature selection significantly improves sample trading strategy results in terms of the overall rate of return, as well as maximum drawdown during a trading period.
High-performance Concrete Compressive Strength Prediction using Time-Weighted Evolutionary Fuzzy Support Vector Machines Inference Model
2012, Automation in Construction
Citation Excerpt :
Thus, estimation requires development of a more advanced time series prediction algorithm, such as that achieved using an AI approach. Refenes et al. [36] described structural change as a time series data characteristic that should always be taken into account in all methodological approaches to time-series analysis. In light of this characteristic, Cao et al. [8] expressed that recent data provide more relevant information than distant data.
The major different between High Performance Concrete (HPC) and conventional concrete is essentially the use of mineral and chemical admixture. These two admixtures made HPC mechanical behavior act differently compare to conventional concrete at microstructures level. Certain properties of HPC are not fully understood since the relationship between ingredients and concrete properties is highly nonlinear. Therefore, predicting HPC behavior is relatively difficult compared to predicting conventional concrete behavior. This paper proposes an Artificial Intelligence hybrid system to predict HPC compressive strength that fuses Fuzzy Logic (FL), weighted Support Vector Machines (wSVM) and fast messy genetic algorithms (fmGA) into an Evolutionary Fuzzy Support Vector Machine Inference Model for Time Series Data (EFSIMT). Validation results show that the EFSIMT achieves higher performance in comparison with Support Vector Machines (SVM) and obtains results comparable with Back-Propagation Neural Network (BPN). Hence, EFSIMT offers strong potential as a valuable predictive tool for HPC compressive strength.
Evolutionary fuzzy decision model for cash flow prediction using time-dependent support vector machines
2011, International Journal of Project Management
The ability of project managers to make reliable cash flow predictions enhances project cost flow control and management. Reliable cash flow prediction over the course of a construction project puts the project manager in a better position to identify potential problems and develop appropriate strategies to mitigate the negative effects of such on overall project success. Therefore, managers should monitor project progress using cash flow data, which has unique characteristics, as time series data. However, the complex, mutable nature of construction projects currently requires significant reliance on experience and expert opinions to predict cash flow on an ongoing basis. Recent studies have indicated good potential for using artificial intelligence to reduce reliance on human input in cash flow prediction processes. The Evolutionary Fuzzy Support Vector Machine Inference Model for Time Series Data (EFSIM_T), an artificial intelligence hybrid system focusing on the management of time series data characteristics which fuses fuzzy logic (FL), weighted support vector machines (weighted SVMs) and a fast messy genetic algorithm (fmGA), represents a promising alternative approach to predicting cash flow. Simulations performed on historical cash flow data demonstrate the EFSIM_T is an effective tool for predicting cash flow.
Flexible neural trees ensemble for stock index modeling
2007, Neurocomputing
Citation Excerpt :
By highlighting the advantages and overcoming the limitations of both the neural networks technique and rule-based systems technique, the hybrid approach can facilitate the development of more reliable intelligent systems to model expert thinking and to support the decision-making processes. Refenes et al. [24] proposed a simple modification to the error backpropagation procedure which takes into account gradually changing input–output relations. The procedure is based on the principle of discounted least squares whereby learning is biased towards more recent observations with long term effects experiencing exponential decay through time.
The use of intelligent systems for stock market predictions has been widely established. In this paper, we investigate how the seemingly chaotic behavior of stock markets could be well represented using flexible neural tree (FNT) ensemble technique. We considered the Nasdaq-100 index of Nasdaq Stock Market^SM and the S&P CNX NIFTY stock index. We analyzed 7-year Nasdaq-100 main index values and 4-year NIFTY index values. This paper investigates the development of novel reliable and efficient techniques to model the seemingly chaotic behavior of stock markets. The structure and parameters of FNT are optimized using genetic programming (GP) like tree structure-based evolutionary algorithm and particle swarm optimization (PSO) algorithms, respectively. A good ensemble model is formulated by the local weighted polynomial regression (LWPR). This paper investigates whether the proposed method can provide the required level of performance, which is sufficiently good and robust so as to provide a reliable forecast model for stock market indices. Experimental results show that the model considered could represent the stock indices behavior very accurately.
Enhanced time-dependent evolutionary fuzzy support vector machines inference model for cash flow prediction and estimate at completion
2013, International Journal of Information Technology and Decision Making

View all citing articles on Scopus

^☆: This research is supported by the Department of Trade and Industry under the NCTT programme and the corporate members of the NeuroForecasting Club. We would like to thank Barclays-BZW, Citibank International, Mars Group, Postel Investment Management, Sabre Fund Management, and Societe Generale, for their material and technical support.

View full text

PaperFinancial time series modelling with discounted least squares backpropagation☆

Abstract

Learning with a slowly changing distribution

Forecasting by smoothed regression

Paper
Financial time series modelling with discounted least squares backpropagation☆