Novel hybrid approach to data-packet-flow prediction for improving network traffic analysis

doi:10.1016/j.asoc.2009.03.003

Applied Soft Computing

Volume 9, Issue 3, June 2009, Pages 1177-1183

https://doi.org/10.1016/j.asoc.2009.03.003 Get rights and content

Abstract

Forecast of the flow of data packets between client and server for a network traffic analysis is viewed as a part of web analytics. Thousands of web-smart businesses depend on web analytics to improve website conversions, reduce marketing costs, facilitate website optimization, speed-up website monitoring and provide a higher level of service to their customers and partners. This paper particularly intends to develop a high accurate prediction as one of core component of network traffic analysis. In this study, a novel hybrid approach, combining adaptive neuro-fuzzy inference system (ANFIS) with nonlinear generalized autoregressive conditional heteroscedasticity (NGARCH), is tuned optimally by quantum minimization (QM) and then applied to forecasting the flow of data packets around website. The composite model (QM-ANFIS/NGARCH) is setup in the forecast point of view to improve the predictive accuracy because it can resolve the problems of the overshoot and volatility clustering simultaneously within time series. As part of real-time intelligence web analytics, the high accurate prediction will aid webmaster to improve the throughput of data-packet-flow up to around 20%, with helping each webmaster to optimize their website, maximize online marketing conversions, and lead campaign tracking.

Introduction

Webmaster in fact does not want to spend money or time on a website that sits idle not yielding an enquiry or a sale. In other words, webmaster want apply website analytics (or free counter) to design websites that are integrated with effective search engine marketing strategies to generate traffic and convert that traffic to sales [1]. What we need is to seek the website analytics that provides detailed return-on-investment analysis for an unlimited number of search engine advertising, banner advertising, affiliate marketing or email marketing campaigns and click in- and out-tracking, combined with website statistics [2]. Therefore, network traffic analysis has become the trusted standard task in website statistics for various internet companies such as travel, dating sites and online shops. This is because website tracking capacity will help each webmaster to optimize their website, maximize online marketing conversions and lead campaign tracking. However, the website tracking or network traffic analysis is closely related to the flow of data packets between hosts. Particularly, a look-ahead prediction applied to forecasting the flow of data packets at upcoming time interval is considered as a measure to foresee the possibility of any fluctuation over flow instantly.

The researches on predictability analysis of network traffic [3], [4] have argued that prediction models with short-range-dependent behavior can capture statistics of (self-similar) traffic quite accurately for the limited time scales of interests in measurement-based traffic management. Those studies [3], [4] also found that the applicability of traffic prediction is limited by the deteriorating prediction accuracy with increasing prediction interval. This implies that a measure from short-term non-periodic forecast with limited time scales could effectively work out a reliable network traffic prediction. Furthermore, a traffic model called fractional autoregressive integrated moving-average (FARIMA) has been proposed for analyzing long-range or short-range dependency over network traffic trace [5] as result of fitting FARIMA to actual traffic trace very well. However, it will take a lot of computation time for fitting traffic trace due to its complexity for modeling a structure of FARIMA [6]. In essence, an improved approach about modeling FARIMA [6] has the feature of transferring the FARIMA problem into an ARMA problem so that any algorithm (not necessarily related to the method used for fitting the ARMA model) could be used to determine the Hurst parameter. This approach first determines the Hurst parameter (differencing level d) before fitting ARMA model, thus reducing the number of plausible models to be examined and the time to identify model iteratively. Accordingly, one can reduce the time to build traffic models and to do real-time modeling. In other words, ARMA model in fact can substitute original FARIMA for simulating actual traffic trace with less computational burden.

Several well-known prediction models have challenged a few crucial problems. For example, grey model (GM) [7] has encountered the overshoot problem such that it will induce big residual errors around turning-point region in time series during the prediction. Auto-regressive moving-average (ARMA) [8] cannot fit data sequences very well for irregular or non-periodic time series due to the lack of a dynamic learning mechanism, back-propagation artificial neural network (BPANN) [9] may have trouble with the problem of over-fitting or under-fitting results due to inappropriate parameters (weights) chosen after training, and adaptive neuro-fuzzy inference system (ANFIS) [10] model cannot avoid the volatility clustering [11] and thus this effect deteriorates the predictive accuracy a lot for the non-periodic short-term prediction. It is noted that ANFIS can possibly overcome the problem of over-fit and automatically tuned/trained by itself. Therefore, in this paper, incorporating a nonlinear generalized autoregressive conditional heteroscedasticity (NGARCH) [12] model into ANFIS system is schemed to tackle the overshoot and volatility clustering effects at the same time during single-step-look-ahead prediction. The proposed composite model ANFIS/NAGRCH is tuned optimally by quantum minimization (QM) [13] to form a linear combination of both models in such a way that it is not only simplify the complex system practically, but also improve the predictive accuracy significantly because of resolving the problems of the overshoot and volatility clustering simultaneously. In short, in order to manage the web resources effectively and efficiently, a high accurate prediction is applicable to foreseeing the in-flow and out-flow data packets to act as part of real-time intelligence network traffic analysis. Web traffic analysis provides valuable information for website administrators to customize the information that is hosted on their web servers so as to reach a larger audience. In the following sections, a novel hybrid (ANFIS/NGARCH) applied to forecasting time series is introduced first to act as one of core component of network traffic analysis. Next, quantum minimization is used to optimally tune the composite model ANFIS/NGARCH. Third, two experiments are demonstrated to verify the high accurate prediction implemented by the one we proposed among inflow and outflow of data packets around websites. Finally, the comparison of two aspects of website traffic monitoring is made with and without foreseeing the packet-flow around websites.

Section snippets

Volatility clustering problem

A certain characteristic commonly associated with financial time series is called volatility clustering, in which large changes tend to follow large changes, and small changes tend to follow small changes. In either case, the changes from one period to the next are typically of unpredictable sign. Volatility clustering, or persistence, suggests a time series model where successive disturbances, even if uncorrelated, are yet serially dependent. Thus large disturbances, either positive or

Quantum minimization—an optimum search algorithm

In the quantum-based representation [13], the operator | 〉 is defined as Dirac's conventional “ket” notation used for the qbits ψ_i such that a quantum state ψ is represented by superposed states $| ψ 〉 = \sum_{i = 1}^{N} ω_{i} | ψ_{i} 〉$ where the probability amplitude ω_i must satisfy $\sum_{i = 1}^{N} | | ω_{i} | |^{2} = 1$ . Quantum-based minimization that makes optimization task work out associated with probability of success at least 1/2 within an unsorted database is realized by quantum minimum searching algorithm [20]. A quantum exponential

QM tuning composite model ANFIS/NGARCH

A single-step-look-ahead prediction can be implemented by adding a variation to the current observed datum [10], and the variation is defined as backward–difference as follows. $\hat{o} (k + 1) = o (k) + δ \hat{o} (k + 1)$ $δ \hat{o} (k + 1) = f (o (k), o (k - 1), \dots, o (k - s), δ o (k), δ o (k - 1), \dots, δ o (k - s))$ $δ o (k) = o (k) - o (k - 1)$ where $\hat{o} (k + 1)$ , o(k), $δ \hat{o} (k + 1)$ , and δo(k) stand for the predicted output at next period, the current true observed datum, the predicted variation at next period, and the current variation, respectively.

We formulate a function of

Criteria for measuring accuracy

A critical point in time series prediction is what criterion can be used to measure the accuracy of the predicted results. Because of the wide applications of time series prediction, the question of whether accuracy is “good enough” is largely dependent on user-specified criteria. Measurement of forecasting performance is highly dependent on how rigidly the criteria are specified for measurement of the degree of accuracy. In order to justify reasonable accuracy for a time series prediction, two

Conclusions

Network traffic analysis provides valuable information for website administrators to customize the information that is hosted on their web servers so as to reach a larger audience. Thus, in order to manage the web resources effectively and efficiently, a high accurate prediction is applicable to foreseeing the inflow and outflow data packets to act as part of real-time intelligence network traffic analysis In this paper, A novel hybrid (QM-ANFIS/NGARCH) is setup in the forecast point of view to

Acknowledgements

This work is fully supported by the National Science Council, Taiwan, Republic of China, under grant number NSC 93-2218-E-143-001.

References (27)

A. Sang et al.
A predictability analysis of network traffic
Computer Networks
(2002)
L. Hentschel
All in the family: nesting symmetric and asymmetric GARCH models
Journal of Financial Economics
(1995)
T.A. Funkhouser et al.
Management of large amounts of data in interactive building walkthroughs
S. Aissi et al.
E-business process modeling: the next big step
IEEE Computer
(2002)
T. Tuan et al.
Congestion Control for Self-Similar Network Traffic, Self-Similar Network Traffic and Performance Evaluation
(2000)
J.K. Liu et al.
Traffic modeling based on FARIMA models
Y.-T. Shu et al.
Internet traffic modeling and prediction using FARIMA models
Chinese Journal of Computers
(2001)
B.R. Chang
Hybrid BPNN-weighted grey-CLMS forecasting
Journal of Information Science and Engineering
(2005)
G.E.P. Box et al.
Time Series Analysis: Forecasting & Control
(1994)
S. Haykin
Neural Networks
(1999)

J.-S.R. Jang

ANFIS: adaptive-network-based fuzzy inference systems

IEEE Transactions on Systems, Man, and Cybernetics

(1993)

C. Gourieroux et al.

Financial Applications

(1997)

T. Bellerslve

Generalized autoregressive conditional heteroscedasticity

Journal of Econometrics

(1986)

Cited by (27)

A novel hybrid model combining βSARMA and LSTM for time series forecasting
2023, Applied Soft Computing
Time series forecasting is an important and active research area due to the significance of prediction and decision-making in several applications. Most commonly used models for time series forecasting are based on Gaussianity assumption, e.g., AR (Autoregressive), ARMA (Autoregressive moving average), ARIMA (autoregressive integrated moving average), etc. But for many applications, the gaussianity presumption is too restrictive, hence non-gaussian based time series models are becoming more and more popular. Many of the hybrid models that are currently used in the literature combine ARIMA and artificial neural network (ANN) while taking various time series data into account with different approaches. Although the accuracy of the predictions made by these models is higher than that of the individual models, there is room for further accuracy improvement if the dynamics of the provided time series is taken into consideration while applying the models. In this study, a new hybrid $β S A R M A - L S T M$ model for time series forecasting is proposed. It combines a non-gaussian based time series model called Beta seasonal autoregressive moving average $β S A R M A$ with a recurrent neural network model called Long Short Term Memory Network (LSTM). The advantage of the proposed model is that $β S A R M A$ is based on the beta distribution, which contains stochastic seasonal dynamics, and LSTM is a recurrent neural network which can be used to a variety of sequential data with high levels of accuracy. In this work, a $β S A R M A$ model is applied on a given time series data in order to identify the linear structure in the data and the error between the original and $β S A R M A$ predicted data is considered as a nonlinear model, which is then modelled using LSTM. The asymptotic stability of the proposed approach is analysed to ensure that the proposed model may not show increasing variance over time. The proposed hybrid $β S A R M A - L S T M$ model along with individual ARIMA, $β S A R M A$ , LSTM, Multilayer perceptron (MLP), and some existing hybrid model ARIMA-ANN was applied on some real, simulated, and experimental datasets such as: relative humidity data, Air passengers, Bitcoin, Sunspots and Mackey Glass series. The results obtained using proposed model for all these data sets show higher prediction accuracy for both one-step and multi-step ahead forecasts.
Framework based on multiplicative error and residual analysis to forecast bitcoin intraday-volatility
2022, Physica A: Statistical Mechanics and its Applications
Citation Excerpt :
These present advantages in comparison with HAR, GARCH, and MEM models since ANNs capture the non-linearity of the series and do not require the series to be stationary for modeling. These reasons have led ANN models to gain popularity for forecasting, being applied to different types of time series ([19; 20; 21; 22; 23; 24; 25; 26]). In all cases, ANNs have shown superior forecasting capabilities to linear models.
This paper deals with bitcoin volatility forecasting, through a framework based on Long-Short Term Memory, multiplicative error and residual analysis. This framework is based on modeling the nonlinear behavior of the multiplicative error in order to improve forecast accuracy. The rapid growth of the cryptocurrency market, its high volatility and its applications in different commercial transactions have attracted the attention of academics and investors. Among cryptocurrencies, bitcoin is the one with the highest volume of trading. For this reason, it is important for investors and companies that use bitcoin as a means of payment to be able to predict its volatility more accurately. Nonetheless, this presents a great challenge given the chaotic behavior of the bitcoin volatility series. The results indicate that the proposed model is able to better capture non-linear behavior of the volatility time series, allowing a more accurate forecast in terms of MSE.
Multivariate intuitionistic fuzzy inference system for stock market prediction: The cases of Istanbul and Taiwan
2022, Applied Soft Computing
Citation Excerpt :
A fusion ANFIS has been proposed for prediction Taiwan stock price series [8]. A hybrid IS, created as a combination of generalized autoregressive conditional heteroscedasticity and ANFIS, has modelled the flow of data packets around the website [35]. ANFISs have been presented for the purpose of prediction short-term trends of the stock exchange [5], and natural gas demand [36].
Many of decision-making and policy planning processes involve a time-series prediction problem and so this area has extensive literature including a great variety of time-series prediction tools and inferences systems. An important part of these is based on fuzzy sets. However, it is known that fuzzy sets may fail to satisfy or characterize the uncertainty of the data in a comprehensive manner because they cannot depict the neutrality degree of time-series. Another important and decisive deficiency of current inference systems is to based on the univariate structure. However, the time series dealt with in a prediction problem generally interact with other time series. Considering these issues, creating an inference system based on intuitionistic fuzzy sets and multivariate relationships for a time series prediction problem is a requirement even an obligation. With these regards, this study presents a multivariate intuitionistic fuzzy time-series definition and its prediction models and introduces a multivariate intuitionistic fuzzy inference system (M-IFIS). The basic novelty of the article can be expressed as the definition of a multivariate intuitionistic fuzzy time series, as well as the creation of a relevant analysis mechanism, first-time in the literature. Sigma-pi neural network is used as an inference tool in M-IFIS and membership and non-membership values and lagged crisp observations of multivariable time-series are used as inputs of it. In order to reveal the performance of the proposed system, Istanbul Stock Exchange (IEX) and Taiwan Stock Exchange Capitalization Weighted Stock Index (TAIEX) are analysed and the results are evaluated as comprehensive and comparative. All findings reveal the superiority M-IFIS in predictive accuracy.
Improving forecasting accuracy of time series data using a new ARIMA-ANN hybrid method and empirical mode decomposition
2019, Neurocomputing
Citation Excerpt :
Therefore, many real world time series data presents complex nonlinear patterns which might not be modeled by ARIMA effectively. For the nonlinear time series modeling, Artificial Neural Networks (ANNs) are one of the most widely used algorithms [6] in many fields, such as finance [7], energy [8], hydrology [9], and network communications [10]. ANNs have several advantages over ARIMA and other forecasting models.
Many applications in different domains produce large amount of time series data. Making accurate forecasting is critical for many decision makers. Various time series forecasting methods exist that use linear and nonlinear models separately or combination of both. Studies show that combining of linear and nonlinear models can be effective to improve forecasting performance. However, some assumptions that those existing methods make, might restrict their performance in certain situations. We provide a new Autoregressive Integrated Moving Average (ARIMA)-Artificial Neural Network (ANN) hybrid method that work in a more general framework. Experimental results show that strategies for decomposing the original data and for combining linear and nonlinear models throughout the hybridization process are key factors in the forecasting performance of the methods. By using these findings, the proposed hybrid method is combined with Empirical Mode Decomposition (EMD) technique which generates more predictable components. We show that our hybrid method with EMD can be an effective way to improve forecasting accuracy obtained by traditional hybrid methods and also any of the individual methods that we used separately.
Intuitionistic time series fuzzy inference system
2019, Engineering Applications of Artificial Intelligence
Citation Excerpt :
A fusion ANFIS procedure has been proposed for the forecasting stock price problems in Taiwan (Cheng et al., 2009). It has been intended to develop a novel hybrid approach combining ANFIS with nonlinear generalized autoregressive conditional heteroscedasticity for the prediction flow of data packets around website (Chang and Tsai, 2009). An ANFIS controller has been used to forecast stock market short-term trends (Atsalakis and Valavanis, 2009).
Although adaptive network fuzzy inference system and fuzzy functions approach can be utilized as a prediction tool, they have been not designed for prediction problem and they ignore the dependency structure of time series observations. From this viewpoint, making a design of the method that considers the dependency structure of observations will provide to get more accurate prediction. In this study, an intuitionistic time series fuzzy inference system (I-TSFIS) has been proposed. In the I-TSFIS, in just the same way as in the intuitionistic fuzzy inference systems, not only the membership values and crisp observations but also the non-membership values are used as inputs. Moreover, due to the use of crisp values as targets and outputs, the output does not need to be deffuzzified. Non-linear relationships between inputs and outputs of the proposed I-TSFIS are determined by Sigma-Pi neural network (SP-NN). The obtaining of optimal weights of SP-NN is performed by modified particle swarm optimization. And also I-TSFIS uses intuitionistic fuzzy C-means to obtain fuzzy clusters, membership and non-membership values of observations for these clusters. To evaluate the prediction performance of the proposed I-TSFIS, various real-life time series data sets have been analyzed and the results demonstrate the superior prediction ability of the proposed I-TSFIS.
A moving-average filter based hybrid ARIMA-ANN model for forecasting time series data
2014, Applied Soft Computing Journal
Citation Excerpt :
For these reasons, ANN models have become more popular in forecasting. ANNs have been applied to various time series data, such as electricity demand data [9], financial data [10], river flow data [11], and network data [12], for forecasting. In all these cases ANNs were shown to yield good forecasts compared with ARIMA models.
A suitable combination of linear and nonlinear models provides a more accurate prediction model than an individual linear or nonlinear model for forecasting time series data originating from various applications. The linear autoregressive integrated moving average (ARIMA) and nonlinear artificial neural network (ANN) models are explored in this paper to devise a new hybrid ARIMA–ANN model for the prediction of time series data. Many of the hybrid ARIMA–ANN models which exist in the literature apply an ARIMA model to given time series data, consider the error between the original and the ARIMA-predicted data as a nonlinear component, and model it using an ANN in different ways. Though these models give predictions with higher accuracy than the individual models, there is scope for further improvement in the accuracy if the nature of the given time series is taken into account before applying the models. In the work described in this paper, the nature of volatility was explored using a moving-average filter, and then an ARIMA and an ANN model were suitably applied. Using a simulated data set and experimental data sets such as sunspot data, electricity price data, and stock market data, the proposed hybrid ARIMA–ANN model was applied along with individual ARIMA and ANN models and some existing hybrid ARIMA–ANN models. The results obtained from all of these data sets show that for both one-step-ahead and multistep-ahead forecasts, the proposed hybrid model has higher prediction accuracy.

View all citing articles on Scopus

¹: Tel.: +886 7 6158000x3315; fax: +886 7 6158001.

View full text

Novel hybrid approach to data-packet-flow prediction for improving network traffic analysis

Abstract

Introduction

Section snippets

Volatility clustering problem

Quantum minimization—an optimum search algorithm

QM tuning composite model ANFIS/NGARCH

Criteria for measuring accuracy

Conclusions

Acknowledgements

Computer Networks

Journal of Financial Economics

Management of large amounts of data in interactive building walkthroughs

E-business process modeling: the next big step

IEEE Computer

Congestion Control for Self-Similar Network Traffic, Self-Similar Network Traffic and Performance Evaluation

Traffic modeling based on FARIMA models

Internet traffic modeling and prediction using FARIMA models

Chinese Journal of Computers

Hybrid BPNN-weighted grey-CLMS forecasting

Journal of Information Science and Engineering

Time Series Analysis: Forecasting & Control

Neural Networks

ANFIS: adaptive-network-based fuzzy inference systems

IEEE Transactions on Systems, Man, and Cybernetics

Financial Applications

Generalized autoregressive conditional heteroscedasticity

Journal of Econometrics