Case Studies and Metrics
Multilabel classification techniques have been applied in many real-world situations in the last two decades. Each one represents a different case study for MLC, using one or more MLDs. After the general overview provided in Sect. 3.1 , this chapter begin
- PDF / 470,028 Bytes
- 31 Pages / 439.37 x 666.142 pts Page_size
- 3 Downloads / 233 Views
Case Studies and Metrics
Abstract Multilabel classification techniques have been applied in many real-world situations in the last two decades. Each one represents a different case study for MLC, using one or more MLDs. After the general overview provided in Sect. 3.1, this chapter begins by briefly describing in Sect. 3.2 the most usual case studies found in the literature. As a result, a full list of available MLDs will be obtained, and the usual characterization metrics are explained and put in use with them in Sect. 3.3. Then, a practical use case is detailed in Sect. 3.4, running a simple MLC algorithm over a few MLDs. Lastly, the usual performance evaluation metrics for MLC are introduced in Sect. 3.5 and they are used to analyze the results obtained from this experiment.
3.1 Overview The main application fields of MLC were introduced in the previous chapter from a global perspective. The goal in this chapter was to delve into each one of these fields, enumerating every one of the publicly available MLDs and stating where they come from. In addition to this basic reference information, it would be interesting to get some general characteristics for each MLD. For doing so, most of the characterization metrics described in the literature are going to be introduced, along with their formulations and discussion about their usefulness. Several extensive tables containing each measurement for every MLD will be provided. In the following chapters, several dozens of MLC algorithms will be described, and some of them will be experimentally tested. Therefore, how to conduct such an experiment, and the way the results can be assessed to evaluate the algorithms’ performance, are fundamental aspects. Once the available MLDs and their main traits are known, a basic kNN-based MLC algorithm is introduced and it is run to process some of these MLDs. Multilabel predictive performance evaluation metrics have to deal with the presence of multiple outputs, taking into consideration the existence of predictions which are partially correct or wrong. As will be expounded, these metrics can be grouped
© Springer International Publishing Switzerland 2016 F. Herrera et al., Multilabel Classification, DOI 10.1007/978-3-319-41111-8_3
33
34
3 Case Studies and Metrics
into several categories according to distinct criteria. Then, most of the MLC evaluation metrics are explained along with their formulation, using them to assess the results obtained from the previous experiments.
3.2 Case Studies In the previous chapter, the main application fields for MLC were portrayed. Attending to the grouping criterion then established, in this section most of the case studies found in the specialized literature will be enumerated. Table 3.1 summarizes these case studies, giving their original references and the place they can be downloaded from.1 Some of these case studies have associated several MLDs, whose names and characteristics will be analyzed later. The same MLD can be available in different formats,2 for instance MULAN, MEKA, and KEEL
Data Loading...
