The ROC curve is a valuable tool to compare different classifiers that can assign a score to their predictions. In general, this score can be interpreted as a probability, so it's bounded between 0 and 1. The plane is structured as shown in the following diagram:
The x-axis represents the increasing false positive rate (1 - FPR) also known as 1 - Specificity, defined as follows:
The y-axis represents the true positive rate (TPR) also known as Sensitivity:
The dashed oblique line in the previous graph represents a perfectly random classifier (in a binary scenario, it's equivalent to tossing a fair coin to make every prediction), so all the curves below this threshold perform worse than a random choice, while the ones above it show better performance. Of course, the best classifier has an ROC curve split into the segments...
