An Enterprise System Bus (ESB) is a software which is used to communicate between various mutually interacting software applications in manufacturing plants. ESB performance is very important for the smooth functioning of the system. Any degradation or failure of the ESB results in huge revenue loss due to production discontinuity. Therefore, maintaining ESB in a healthy state is very essential and there are multiple factors related to resource utilization, workload and number of interfaces etc., which influences the performance of the ESB. Forecasting these variables at least a day ahead (24 h ahead) is required to take appropriate actions by the business team to maintain the ESB performance under control. But, these variables are heterogeneous (continuous, discrete and percentages) in nature, highly non-linear and non-stationary. The challenges associated with forecasting of these variables are (i) long horizon (24 h ahead forecast at 5 min granularity requires to forecast 288 steps) (ii) data generated from these kinds of systems makes it very difficult to use any linear statistical methods like state-space models, ARIMA etc. To address these challenges, the paper presents a framework where a basket of learning algorithms based on Artificial Neural Network (ANN), Support Vector Regression (SVR) and Random Forests (RF) were used to model the chaotic behavior of the time series with a real-time automatic algorithm selection mechanism which enables appropriate forecasting algorithm to be chosen dynamically based on the performance over a time window, resulting in different algorithms being used for forecasting the same target variable on different days. Importance of the proposed strategy was demonstrated with suitable forecasting results for different variables/parameters impacting the performance of the critical Enterprise System Bus of an automotive manufacturing setup.
[1]
Geetam Singh Tomar,et al.
The Performance Metric for Enterprise Service Bus (ESB) in SOA system: Theoretical underpinnings and empirical illustrations for information processing
,
2017,
Inf. Syst..
[2]
Amr Badr,et al.
Forecasting of nonlinear time series using ANN
,
2017
.
[3]
Amir F. Atiya,et al.
An Empirical Comparison of Machine Learning Models for Time Series Forecasting
,
2010
.
[4]
J. Stock,et al.
A Comparison of Direct and Iterated Multistep Ar Methods for Forecasting Macroeconomic Time Series
,
2005
.
[5]
Cardona Alzate,et al.
Predicción y selección de variables con bosques aleatorios en presencia de variables correlacionadas
,
2020
.
[6]
John R. Freeman.
Granger Causality and the Time Series Analysis of Political Relationships
,
1983
.
[7]
Grzegorz Dudek,et al.
Short-Term Load Forecasting Using Random Forests
,
2014,
IEEE Conf. on Intelligent Systems.
[8]
Bukola Titilayo Ojemakinde.
Support Vector Regression for Non-Stationary Time Series
,
2006
.
[9]
Antanas Verikas,et al.
Mining data with random forests: A survey and results of new tests
,
2011,
Pattern Recognit..
[10]
Leo Breiman,et al.
Random Forests
,
2001,
Machine Learning.
[11]
Marc-Thomas Schmidt,et al.
The Enterprise Service Bus: Making service-oriented architecture real
,
2005,
IBM Syst. J..
[12]
Georgia Papacharalampous,et al.
Variable Selection in Time Series Forecasting Using Random Forests
,
2017,
Algorithms.
[13]
William Remus,et al.
Neural Network Models for Time Series Forecasts
,
1996
.
[14]
Bernhard Schölkopf,et al.
A tutorial on support vector regression
,
2004,
Stat. Comput..