Interactive machine learning: letting users build classifiers
✍ Scribed by MALCOLM WARE; EIBE FRANK; GEOFFREY HOLMES; MARK HALL; IAN H WITTEN
- Book ID
- 102572097
- Publisher
- Elsevier Science
- Year
- 2001
- Tongue
- English
- Weight
- 315 KB
- Volume
- 55
- Category
- Article
- ISSN
- 1071-5819
No coin nor oath required. For personal study only.
✦ Synopsis
According to standard procedure, building a classi"er using machine learning is a fully automated process that follows the preparation of training data by a domain expert. In contrast, interactive machine learning engages users in actually generating the classi"er themselves. This o!ers a natural way of integrating background knowledge into the modelling stage*as long as interactive tools can be designed that support e$cient and e!ective communication. This paper shows that appropriate techniques can empower users to create models that compete with classi"ers built by state-of-the-art learning algorithms. It demonstrates that users*even users who are not domain experts*can often construct good classi"ers, without any help from a learning algorithm, using a simple two-dimensional visual interface. Experiments on real data demonstrate that, not surprisingly, success hinges on the domain: if a few attributes can support good predictions, users generate accurate classi"ers, whereas domains with many high-order attribute interactions favour standard machine learning techniques. We also present an arti"cial example where domain knowledge allows an &&expert user'' to create a much more accurate model than automatic learning algorithms. These results indicate that our system has the potential to produce highly accurate classi"ers in the hands of a domain expert who has a strong interest in the domain and therefore some insights into how to partition the data. Moreover, small expert-de"ned models o!er the additional advantage that they will generally be more intelligible than those generated by automatic techniques.