- Title
- From ensemble learning to meta-analytics: a review on trends in business applications
- Creator
- Haque, Mohammad Nazmul; Moscato, Pablo
- Relation
- ARC.FT120100060 http://purl.org/au-research/grants/arc/FT120100060
- Relation
- Business and Consumer Analytics: New Ideas p. 703-731
- Publisher Link
- http://dx.doi.org/10.1007/978-3-030-06222-4_18
- Publisher
- Springer Nature
- Resource Type
- book chapter
- Date
- 2019
- Description
- Ensemble learning has been applied in different areas to improve the predictive performances using multiple learners. The two core building blocks, diversity and combination rule, which play a significant role in ensemble learners. The ensemble approach can be divided into two broad groups based on the variation of base classifiers: homogeneous and heterogeneous ensemble. We conducted a comprehensive review of the ensemble learning used for data analytics. The study has proceeded from the feature selection to classification. We found that the ensemble learning helps to overcome the problem associated with the dimensionality and class imbalance of data. For this reason, the ensemble approach found to be more suitable for the classification of high-dimensional data. Then we move towards the meta-analytics using ensemble learning. Our comprehensive review of the metaheuristics-based ensemble learning for homogeneous and heterogeneous ensemble found a substantial number of applications of ensemble learning from these categories. The in detail study of the ensemble learning in business applications able to identify four successful application areas: purchasing and marketing, predictive analytics, business process management, and CustomerchurnpredictionPredictioncustomer churncustomer churn prediction. From these application areas, we observe that majority of the approaches built homogeneous ensembles with dynamic selection for single objective optimization. Despite these success in various application domains, ensemble learning could face challenges in analytics in the future. We concluded the chapter with identifying those difficulties and some trends to overcome them for ensemble learning with meta-analytics.
- Subject
- classifier ensembles; ensemble learning; meta-analytics; failure prediction; multi-objective optimization; machine learning
- Identifier
- http://hdl.handle.net/1959.13/1459096
- Identifier
- uon:45579
- Identifier
- ISBN:9783030062224
- Language
- eng
- Hits: 713
- Visitors: 709
- Downloads: 0
| Thumbnail | File | Description | Size | Format |
|---|