As we have a large number of combinations among the base methods, metrics and datasets, a global analysis of the performance of all these combinations is not an easy task.
The service metrics adopted here have been established as part of the Forest Ecosystem Services Toolkit (FEST); more information on methods, metrics, and datasets can be found on the FEST website at www.forestecoservices.net and in Beier et al. (2015) and Caputo et al. (2016).
Lack of clarity in the interpretation of statistical methods, metrics, and results that document novel and poorly understood structural patterns can only be a detriment to this endeavor.
In general, cluster validity can be assessed on three bases: within method metrics, between-method metrics and cluster significance [ 57].
To quantitatively evaluate the performance of the proposed method, metrics for object space and for boundary precision are, respectively, considered.
McCabe's and Halstead's metrics are called module or method level metrics and the faulty or non-faulty label is assigned to each one of the modules.
Distinct parameters such as method-level metrics and two feature selection approaches which are principal component analysis and correlation based feature selection are used to evaluate the finest performance among the others.
The considered KPIs, prediction methods and metrics, regressors, and sampling methods are summarized in Table 2.
Despite the small differences by individual metrics, Figure 5 shows that these five methods exhibit comparable performance: no method outperforms the remaining methods across all metrics and the maximal difference by any metric is within 0.05. Figure 5(a) and 5(b) show two concave curves of Pearson (Truth) and Spearman (Truth) for RNA-Skim by varying its sig-mer length.
The within-method metrics indicate that Clusters 2 and 5 showed homogeneity and Clusters 3 and 5 showed good separation in terms of average silhouette (Table 3).
