ROCR: Visualizing classifier performance in R

Department of Computational Biology and Applied Algorithmics, Max-Planck-Institute for Informatics, Saarbrücken, Germany.
Bioinformatics (Impact Factor: 4.98). 11/2005; 21(20):3940-1. DOI: 10.1093/bioinformatics/bti623
Source: PubMed


ROCR is a package for evaluating and visualizing the performance of scoring classifiers in the statistical language R. It
features over 25 performance measures that can be freely combined to create two-dimensional performance curves. Standard methods
for investigating trade-offs between specific performance measures are available within a uniform framework, including receiver
operating characteristic (ROC) graphs, precision/recall plots, lift charts and cost curves. ROCR integrates tightly with R's
powerful graphics capabilities, thus allowing for highly adjustable plots. Being equipped with only three commands and reasonable
default values for optional parameters, ROCR combines flexibility with ease of usage.

Availability: ROCR can be used under the terms of the GNU General Public License. Running within R, it is platform-independent.

Contact: tobias.sing{at}

83 Reads
  • Source
    • "We conducted all analyses in program R version 3.0.2 (R Development Core Team 2013), relying on the packages adehabitatHR (Calenge 2006), rgdal (Keitt et al. 2013), rgeos (Bivand and Rundel 2013), sp (Bivand et al. 2008), raster (Hijmans and van Etten 2012), and spatstat (Baddeley and Turner 2005) for spatial analyses, geepack (Halekoh et al. 2006) to implement GEEs, lme4 (Bates et al. 2013) to implement linear mixed-effects models, lmerTest (Kuznetsova et al. 2013) to calculate P-values for linear mixed-effects models, and ROCR (Sing et al. 2005) to calculate ROC scores. "
    [Show abstract] [Hide abstract]
    ABSTRACT: Understanding spatiotemporal variability in prey accessibility is important for disentangling predator-prey interactions and is relevant to management interventions to reduce predation. Recently, caribou (Rangifer tarandus) in Newfoundland declined by 66%, with calf predation by black bears (Ursus americanus) implicated as a major proximate mechanism of the decline. Most predation occurs when calves are aggregated on calving grounds. We used telemetry data from 271 caribou and 45 black bears in 2 caribou herd ranges to examine spatial variability in calf accessibility, identify the distribution of potentially predatory bears, and assess the aggregative response of bears to the calf resource. We predicted whether a bear was a visitor to a calving ground during the calving season (a potentially predatory bear) based upon its sex, the herd range it occupied, its distance to the calving grounds, and the season. The distribution of potentially predatory bears and their degree of segregation from non-predatory bears varied seasonally. The probability of a bear visiting the calving grounds during calving decreased with increasing distance from the calving grounds, and was greater for males than for females in all seasons at distances beyond 2.4 km from the calving grounds. Residency time of bears increased in the calving grounds of 1 herd during calving, suggesting an aggregative response to neonates in that area. For both herds, the estimated distribution of potentially predatory bears was much larger than the calving grounds, illustrating that the relevant scale of predator-prey interactions may extend far beyond the area where lethal encounters occur. Our work highlights the value of examining spatiotemporal dynamics of predator movements prior to implementing ecosystem manipulations designed to reduce predation and provides a modeling framework that can be used to guide management interventions in systems with aggregated prey. © 2015 The Wildlife Society.
    Journal of Wildlife Management 08/2015; DOI:10.1002/jwmg.936 · 1.73 Impact Factor
    • "ances of the investigated classifier ( Fawcett , 2004 ) . The Area Under the ROC curve ( AUC ) was also calculated , to consider in a single scalar value the ability of the classifier to differentiate between the distributions of the two classes ( Hand and Till , 2001 ) . ROC analysis was conducted using ROCR toolkit for the R computing language ( Sing et al . , 2005 ) ."
    [Show abstract] [Hide abstract]
    ABSTRACT: Carcinogenesis is a multi-step process involving genetic alterations and non-genotoxic mechanisms. The in vitro cell transformation assay (CTA) is a promising tool for both genotoxic and non-genotoxic carcinogenesis. CTA relies on the ability of cells (e.g. BALB/c 3T3 mouse embryo fibroblasts) to develop a transformed phenotype after the treatment with suspected carcinogens. The classification of the transformed phenotype is based on coded morphological features, which are scored under a light microscope by trained experts. This procedure is time-consuming and somewhat prone to subjectivity. Herewith we provide a promising approach based on image analysis to support the scoring of malignant foci in BALB/c 3T3 CTA. The image analysis system is a quantitative approach, based on measuring features of malignant foci: dimension, multilayered growth, and invasivity into the surrounding monolayer of non-transformed cells. A logistic regression model was developed to estimate the probability for each focus to be transformed as a function of three statistical image descriptors. The estimated sensitivity of the derived classifier (untransformed against Type III) was 0.9, with an Area Under the Curve (AUC) value equal to 0.90 under the Receiver Operating Characteristics (ROC) curve. Copyright © 2015. Published by Elsevier Ltd.
    Toxicology in Vitro 07/2015; 29(7). DOI:10.1016/j.tiv.2015.07.013 · 2.90 Impact Factor
  • Source
    • "Backward elimination of predictors was then performed using the function Bdrop1,^ until Akaike information criterion (AIC) was minimized. Receiver operating characteristic (ROC) curve analysis was done on the reduced models, using the ROCR package in R (Sing et al. 2005). We interpreted fit values in the reduced models. "
    [Show abstract] [Hide abstract]
    ABSTRACT: As in other highly sexually dimorphic, group-living animals, reproduction in gorillas has been largely viewed as the outcome of competition among males. However, females may exert choice via dispersal decisions or choice of partner in multimale groups, and males may also mate selectively. Here, we examine the paternity of 79 wild mountain gorilla offspring born into four groups characterized by stable dominance hierarchies and the presence of mature offspring of the dominant male. We found that on average the dominant male sires the majority (72 %) of the offspring in stable multimale groups and subordinate males also produce offspring, particularly when dominant males become older or the number of competing males increases. Although expected to disperse to avoid inbreeding, only half of the maturing daughters of dominant males left the group in which their father maintained dominance. However, in all five cases of reproduction by a resident daughter of a dominant male, a subordinate male was the sire of the offspring. As females commonly initiate and end copulations, and dominant males may prefer mating with fully mature females, both male and female mate preferences in addition to male competition apparently play a role in reproductive patterns in multimale groups, emphasizing the complexity of social dynamics in one of our closest living relatives.
    Behavioral Ecology and Sociobiology 07/2015; 69(7). DOI:10.1007/s00265-015-1930-0 · 2.35 Impact Factor
Show more


83 Reads
Available from