An evolutionary algorithm for optimizing the target ordering in Ensemble of Regressor Chains

Moyano, Jose M.; Gibaja, Eva; Ventura, Sebastián

doi:10.1109/cec.2017.7969548

Cited by 16 publications

(14 citation statements)

References 12 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Firstly, the possibility of using any base regressor, or even a hybrid set of regressors, could lead to better prediction performance and exploration of a particular problem characteristics. Besides that, problem adaptation methods improve the resulting model modularity and conceptual simplicity, presenting better accuracy than state-of-the-art methods [21,22,25].…”

Section: Related Workmentioning

confidence: 99%

“…Aiming at improving the ERC algorithm, two localbased MTR methods were proposed recently. Moyano et al [22] proposed the use of Genetic Algorithm (GA) to find an improved set of target chains and to explore the linear correlation among target variables. Their method performs better than ERC in some scenarios, but solving the combinatory problem brings a new processing overhead to the already very costly ERC algorithm.…”

Section: Related Workmentioning

confidence: 99%

“…It is clear that a metric should be employed when describing dependencies between target variables and constructing the CTs. Some recent works used the linear correlation coefficient to explain target dependencies [21,22]. Other measures could be employed for the same purpose, leading to different representations and predictive performances.…”

Section: Multi-output Tree Chainingmentioning

confidence: 99%

“…RRMSE measures a model error decrease over a naive predictor which always outputs the target mean value. The latter servers a baseline in the metric and allows the measurement of the improvement over a shallow predictor, being used in various MTR works [2,[20][21][22][23]25] to compare non-homogeneous targets distributions. The aRRMSE calculation is presented in Eq.…”

Section: Performance Evaluationmentioning

confidence: 99%

“…Several MTR methods belonging to both approaches were proposed in past years [1,2,17,21,22,25]. Some of the local approaches had achieved the highest performance in multiple MT scenarios, outperforming global methods [21,22,25,26]. However, the actual state-of-art methods present some limitations in the interpretation of targets dependencies and have high computational cost solutions.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Multi-Output Tree Chaining: An Interpretative Modelling and Lightweight Multi-Target Approach

Mastelini

Costa

Santana

et al. 2018

J Sign Process Syst

View full text Add to dashboard Cite

Multi-target regression (MTR) regards predictive problems with multiple numerical targets. To solve this, machine learning techniques can model solutions treating each target as a separated problem based only on the input features. Nonetheless, modelling inter-target correlation can improve predictive performance. When performing MTR tasks using the statistical dependencies of targets, several approaches put aside the evaluation of each pair-wise correlation between those targets, which may differ for each problem. Besides that, one of the main drawbacks of the current leading MTR method is its high memory cost. In this paper, we propose a novel MTR method called Multi-output Tree Chaining (MOTC) to overcome the mentioned disadvantages. Our method provides an interpretative internal tree-based structure which represents the relationships between targets denominated Chaining Trees (CT). Different from the current techniques, we compute the outputs dependencies, one-by-one, based on the Random Forest importance metric. Furthermore, we proposed a memory friendly approach which reduces the number of required regression models when compared to a leading method, reducing computational cost. We compared the proposed algorithm against three MTR methods (Single-target-ST; Multi-Target Regressor Stacking-MTRS; and Ensemble of Regressor Chains-ERC) on 18 benchmark datasets with two base regression algorithms (Random Forest and Support Vector Regression). The obtained results show that our method is superior to the ST approach regarding predictive performance, whereas, having no significant difference from ERC and MTRS. Moreover, the interpretative tree-based structures built by MOTC pose as great insight on the relationships among targets. Lastly, the proposed solution used significantly less memory than ERC being very similar in predictive performance.

show abstract

Section: Related Workmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

Section: Multi-output Tree Chainingmentioning

confidence: 99%