A dynamic and on-line ensemble regression for changing environments

Soares, Symone; Araújo, Rui

doi:10.1016/j.eswa.2014.11.053

Cited by 47 publications

(31 citation statements)

References 31 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The more reasonable method is to equip the local models with dynamic weights assignment according to the correlation between query data and local regions, which indicates that adaptive weighting manner is preferable. To this end, the adaptive weights are generally assigned as the normalized distances from query data to the centers of each local region or the quantified prediction performance of the latest measured sample . However, both distance‐based and prediction performance–based approaches are confronted with potential problems.…”

Section: Introductionmentioning

confidence: 99%

Adaptive soft sensor method based on online selective ensemble of partial least squares for quality prediction of chemical process

Cang

Yang

2019

Asia-Pacific J Chem Eng

View full text Add to dashboard Cite

For accurate state prediction of chemical process characterized by connatural features of non-linearity and time variation, a novel online selective ensemble partial least squares learning paradigm, referred to as OSE-PLS, is proposed in this paper. First, the process states are differentiated to diverse local regions by just-in-time learning-based localization method to handle process non-linearity.Then, support vector domain description algorithm is applied to identify and remove redundant ones. Consequently, a variety of local PLS models are established for representing diverse process states. Inspired by the local outlier probability algorithm in outlier detection field, the combination weights of each PLS model are adaptively distinguished by definitely quantifying the probability that query data belong to an outlier in each local region. Finally, the prediction for query data is achieved through selective ensemble learning strategy. To effectively address the inherent time-varying characteristics of process, the OSE-PLS technique is equipped with adaptation mechanism of local model updating and online local model extraction, which enables OSE-PLS to acquire the capability of addressing both gradual and abrupt changes in process simultaneously. An industrial bisphenol-A distillation process is employed to demonstrate the superiorities of novel online selective ensemble partial least squares (OSE-PLS) approach. KEYWORDS adaptive soft sensor, just-in-time learning, local outlier probability, selective ensemble learning, support vector domain description Asia-Pac J Chem Eng. 2019;14:e2346.wileyonlinelibrary.com/journal/apj

show abstract

Section: Introductionmentioning

confidence: 99%

Adaptive soft sensor method based on online selective ensemble of partial least squares for quality prediction of chemical process

Cang

Yang

2019

Asia-Pacific J Chem Eng

View full text Add to dashboard Cite

show abstract

“…• Ensemble pruning -In practice, the ensemble size is usually bounded due to the limitation of resources. A simple pruning strategy is to remove the worst performance model whenever the upper bound of the ensemble is reached [22,23]. The effective ensemble size can also be dynamically determined by approaches, such as instance based pruning [17] and ordered aggregation [24].…”

Section: Introductionmentioning

confidence: 99%

“…The major difference between our work and some previous efforts for power plant modeling [28] is that we consider modeling in dynamic environments. The algorithm we applied is mostly based on the DOER algorithm (Dynamic and On-line Ensemble Regression) proposed in [23], considering its overall better performance on multiple synthetic and real (industry applications) data sets when compared to several state-of-the-art algorithms. For the same reason, we will not provide comparison to other approaches in this paper.…”

Section: Introductionmentioning

confidence: 99%

Power plant performance modeling with concept drift

Xu¹,

Xu²,

Yan³

2017

2017 International Joint Conference on Neural Networks (IJCNN)

View full text Add to dashboard Cite

Abstract-Data-driven predictive analytics are in use today across a number of industrial applications, but further integration is hindered by the requirement of similarity among model training and test data distributions. This paper addresses the need of learning from possibly nonstationary data streams, or under concept drift, a commonly seen phenomenon in practical applications. A simple dual-learner ensemble strategy, alternating learners framework, is proposed. A long-memory model learns stable concepts from a long relevant time window, while a shortmemory model learns transient concepts from a small recent window. The difference in prediction performance of these two models is monitored and induces an alternating policy to select, update and reset the two models. The method features an online updating mechanism to maintain the ensemble accuracy, and a concept-dependent trigger to focus on relevant data. Through empirical studies the method demonstrates effective tracking and prediction when the steaming data carry abrupt and/or gradual changes.

show abstract

“…However, all these listed methods do not add and remove models over time; but the on-line inclusion and removal of models is an important key for improving ensemble prediction performance. Additive Expert (AddExp) [26], On-line Weighted Ensemble (OWE) [27] and Dynamic and On-line Ensemble Regression (DOER) [28] add new models when the ensemble's error on the newest sample is greater than a threshold, and remove inaccurate models over time; while Online Accuracy Updated Ensemble (OAUE) [29] and Learn ++ .NSE [21] add and remove models when a batch is available.…”

Section: Introductionmentioning

confidence: 99%

An adaptive ensemble of on-line Extreme Learning Machines with variable forgetting factor for dynamic system prediction

Soares

Araújo

2016

Neurocomputing

Self Cite

View full text Add to dashboard Cite

A demand for predictive models for on-line estimation of variables is increasing in industry. As industrial processes are timevarying, on-line learning algorithms should be adaptive to capture process changes. On-line ensemble methods have been shown to provide better generalization performance than single models in changing environments. However, most on-line ensembles do not include and exclude models during on-line operation. As a result, the ensembles have limited adaptation capability. Moreover, a higher performance can be obtained by combining a selected set of most relevant models of the ensemble for the current situation, rather than combining all the models. This paper proposes a new on-line learning ensemble of regressor models using an ordered aggregation (OA) technique which is able to provide on-line predictions of variables in changing environments. OA dynamically selects an optimal size and composition of a subset of models based on the minimization of the ensemble error on the newest sample. The proposed strategy overcomes the problem of defining the optimal ensemble size, and in most cases it obtains better performance than aggregating all the models. Models are added or removed for assuring adaptation of the ensemble in changing environments. Furthermore, this paper proposes and integrates a new on-line Extreme Learning Machine (ELM) neural network model with variable forgetting factor (FF) using the directional FF method which shows superior performance in industrial applications when compared to the well-known On-line Sequential ELM (OS-ELM) algorithm. Experiments are reported to demonstrate the performance and effectiveness of the proposed methods.

show abstract

A dynamic and on-line ensemble regression for changing environments

Cited by 47 publications

References 31 publications

Adaptive soft sensor method based on online selective ensemble of partial least squares for quality prediction of chemical process

Adaptive soft sensor method based on online selective ensemble of partial least squares for quality prediction of chemical process

Power plant performance modeling with concept drift

An adaptive ensemble of on-line Extreme Learning Machines with variable forgetting factor for dynamic system prediction

Contact Info

Product

Resources

About