The Many Faces of ROC Analysis
Posted by Armando Brito Mendes | Filed under materiais para profissionais
clicar na imagem para seguir o link
Bom tutorial sobre curvas ROC
Receiver Operating Characteristics (ROC) Analysis originated from signal detection theory, as a model of how well a receiver is able to detect a signal in the presence of noise. Its key feature is the distinction between hit rate (or true positive rate) and false alarm rate (or false positive rate) as two separate performance measures. ROC analysis has also widely been used in medical data analysis to study the effect of varying the threshold on the numerical outcome of a diagnostic test. It has been introduced to machine learning relatively recently, in response to classification tasks with varying class distributions or misclassification costs (hereafter referred to as skew). ROC analysis is set to cause a paradigm shift in machine learning. Separating performance on classes is almost always a good idea from an analytical perspective. For instance, it can help us to
- understand the behaviour and skew-sensitivity of many machine learning metrics, including rule learning heuristics and decision tree splitting criteria, by plotting their isometrics in ROC space;
- develop new metrics specifically designed to improve the Area Under the ROC Curve (AUC) of a model;
- understand fundamental algorithms such as the separate-and-conquer or sequential covering rule learning algorithm, by tracing its trajectory through a sequence of ROC spaces.
The goal of this tutorial is to develop the ROC perspective in a systematic way, demonstrating the many faces of ROC analysis in machine learning.
Tags: data mining, DW \ BI, machine learning
Comments are closed.
