Download | - View accepted manuscript: How automated machine learning help business data science teams? (PDF, 1.1 MiB)
|
---|
DOI | Resolve DOI: https://doi.org/10.1109/ICMLA.2019.00196 |
---|
Author | Search for: Ebadi, Ashkan1; Search for: Gauthier, Yvan; Search for: Tremblay, Stéphane1; Search for: Paul, Patrick1 |
---|
Affiliation | - National Research Council of Canada. Digital Technologies
|
---|
Format | Text, Article |
---|
Conference | 18th IEEE International Conference on Machine Learning and Applications (ICMLA), Dec. 16-19, 2019, Boca Raton, Florida |
---|
Subject | automated machine learning; augmented analystics; decision making; decision support; data science; industry; artificial intelligence; supervised learning |
---|
Abstract | Artificial intelligence and machine learning have attracted the attention of many commercial and non-profit organizations aiming to leverage advanced analytics, in order to provide a better service to their customers, increase their revenues through creating new or improving their existing internal processes, and better exploit their data by discovering complex hidden patterns. Such advanced solutions require data scientists with rare (and generally expensive) skill sets. Moreover, such solutions are often perceived as complex black boxes to decision-makers. Automated machine learning tools aim to reduce the expertise gap between the technical teams and stakeholders involved in business data science projects, by reducing the amount of time and specialized skills required to generate predictive models. We systematically benchmarked five automated machine learning tools against seven supervised learning problems of a business nature. Our results suggest that such tools, in fully automated mode, must be used cautiously, only where predictive models support low-impact decisions and do not need to be explainable, and only by data scientists capable to ensure that all phases of the data mining process have been performed adequately. |
---|
Publication date | 2019-12-19 |
---|
Publisher | IEEE |
---|
In | |
---|
Language | English |
---|
Peer reviewed | Yes |
---|
Export citation | Export as RIS |
---|
Report a correction | Report a correction (opens in a new tab) |
---|
Record identifier | 4f1e766f-0f8b-4ce0-ac97-48e1f64b2c66 |
---|
Record created | 2020-01-20 |
---|
Record modified | 2020-06-04 |
---|