Evolutionary Automated Feature Engineering

Zhu, Guanghui; Shen, Jian; Guo, Xu; Yuan, Chunfeng; Huang, Yihua

doi:10.1007/978-3-031-20862-1_42

Cited by 3 publications

(11 citation statements)

References 19 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In this section, we compare performance results of automated and traditional feature engineering techniques commonly employed in practice and in the scientific literature. The first set of results (shown in Table 5) show classification accuracies (F-1 scores [292]) obtained from experimental evaluations by several authors -specifically, [59], [278], [288], [293]. The feature engineering methods compared are the following.…”

Section: Performance Of Automated Feature Engineering Methodsmentioning

confidence: 99%

“…In the literature, this approach is commonly referred to as Expansion-Reduction. The rest of the methods are LFE [278], NFS [59], AutoFeat [245] and DiFFER [288]. We selected 23 datasets where performance results are widely available for many of the aforementioned automamted feature engineering approaches.…”

Section: Performance Of Automated Feature Engineering Methodsmentioning

confidence: 99%

“…While individual feature processing tasks can be automating separately, many modern automated feature processing approaches [134], [274], [275], [286], [287], [288] are aimed at performing the full range of feature engineering tasks (i.e., feature extraction, synthesis of additional features, and selection of useful features) in a holistic and end-to-end manner. In this works, the performance of the various subtasks may not be distinctive.…”

Section: End-to-end Automated Feature Engineeringmentioning

confidence: 99%

“…The search is conditioned on maximizing information gain. Evolutionary Automated Feature Engineering (EAAFE) [286] and Ring Theory based Harmony Search (RTHS) [290] employ evolutionary computational algorithms to find useful features. The use of reinforcement learning approaches is also common [276].…”

Section: End-to-end Automated Feature Engineeringmentioning

confidence: 99%

“…Neural feature search (NFS) [59] uses a reinforcement learning-based RNN controllers to learn effective transformation policies for generating useful features. Differentiable Automated Feature Engineering (DIFER) [288] formulates the automated feature engineering task as a differentiable problem by mapping discrete features into a continuous vector space with the help of a so-called encoder-predictordecoder-based feature optimizer and then finds effective features by gradient ascent. The general architecture of their framework is shown in Figure 22.…”

Section: End-to-end Automated Feature Engineeringmentioning

confidence: 99%

See 4 more Smart Citations

Data augmentation: A comprehensive survey of modern approaches

Mumuni

2022

Array

179

View full text Add to dashboard Cite

Section: Performance Of Automated Feature Engineering Methodsmentioning

confidence: 99%

Section: Performance Of Automated Feature Engineering Methodsmentioning

confidence: 99%

Section: End-to-end Automated Feature Engineeringmentioning

confidence: 99%

Section: End-to-end Automated Feature Engineeringmentioning

confidence: 99%

Section: End-to-end Automated Feature Engineeringmentioning

confidence: 99%

See 3 more Smart Citations

Data augmentation: A comprehensive survey of modern approaches

Mumuni

2022

Array

179

View full text Add to dashboard Cite

Automated data processing and feature engineering for deep learning and big data applications: A survey

Mumuni,

Mumuni

2024

Journal of Information and Intelligence

View full text Add to dashboard Cite

Empowering Machine Learning with Scalable Feature Engineering and Interpretable AutoML

Eldeeb,

Elshawi

2024

IEEE Trans. Artif. Intell.

View full text Add to dashboard Cite

Automated feature engineering has gained considerable attention in academia and industry. Nevertheless, existing systems often lack practical scalability and efficiency. This paper introduces BigFeat, a scalable and interpretable framework that streamlines critical phases of the machine learning pipeline: feature engineering, model selection, and hyperparameter tuning. BigFeat presents two execution options: as a standalone feature engineering framework, denoted as BigFeat-FE, and as an AutoML framework, referred to as BigFeat-AutoML. BigFeat-FE optimizes input feature quality with the ultimate aim of maximizing predictive performance, based on a user-defined metric. BigFeat-FE employs a dynamic feature generation and selection mechanism that systematically creates a set of expressive features. These features not only enhance prediction performance but also prioritize interpretability. BigFeat-FE employs a metalearning technique to warm-start the optimization process, resulting in significant overall performance gains. BigFeat-AutoML, tailored for algorithm selection and hyperparameter tuning, harnesses a random search method over the space of interpretable models. We conducted extensive experiments, and the results demonstrate that BigFeat-FE consistently outperforms stateof-the-art automated feature engineering frameworks, such as AutoFeat and SAFE, across a wide range of datasets, achieving an average performance improvement of 8.65% compared to AutoFeat and 4.71% compared to SAFE, respectively. Additionally, BigFeat-AutoML demonstrates substantial performance improvement when compared to TPOT and Autosklearn, with average improvements of 0.74% over TPOT and 2.25% over Autosklearn, respectively. Furthermore, BigFeat's scalability is affirmed through its linear complexity, and execution times, averaging 20 times faster than AutoFeat and 14 times faster than SAFE.Impact Statement-The emergence of automated scalable interpretable feature engineering is reshaping the landscape of data science and machine learning. It automates feature creation and selection, providing interpretable features while eliminating the time-consuming burden of manual feature engineering. By streamlining the feature engineering process, this advancement empowers data scientists to concentrate on higher-level tasks and model development. Its scalability ensures it can handle vast and complex datasets across various industries, from healthcare to finance. Interpretability takes center stage in this innovation, enhancing model trustworthiness and facilitating regulatory compliance. The automation of different stages of the machine learning pipeline, including feature engineering, model selection, and hyperparameter tuning, holds the promise of improving predictive modeling, decision-making, and overall efficiency in data-driven applications. This heralds a new era in data science and machine learning, where the potential for transformative advancements is on the horizon.

show abstract

Evolutionary Automated Feature Engineering

Cited by 3 publications

References 19 publications

Data augmentation: A comprehensive survey of modern approaches

Data augmentation: A comprehensive survey of modern approaches

Automated data processing and feature engineering for deep learning and big data applications: A survey

Empowering Machine Learning with Scalable Feature Engineering and Interpretable AutoML

Contact Info

Product

Resources

About