DrawSth is a user friendly package, which aims to make statistic graphs and calculate relative evaluation metrics of classifiers.
It's capable of drawing four different charts including receiver operating characteristic (ROC) curve, precision-recall (PR) curve, Enrichment curve and log-scaled area under the receiver-operating characteristic (logAUC) curve.
The ROC curve reflects the relationship between the true and false positive rates,The AUC is commonly used in machine learning studies as an accuracy metric and is defined as the area under the ROC curve. The ROC curve was plotted for each compound in prediction order. The possible values of the AUC range from 0 to 1, and the AUC values corresponding to ideal and random prediction are 1 and 0.5, respectively.
Similar to the ROC curve, P-R curve is introduced to characterize of precision and recall on the whole sample set. The area under the P-R curve indicates the possibility of the model to achieve high precision and recall rate at the same time to a certain extent. Therefor, the larger the AUC of P-R curve, the better the model performs.
xxxxx
LogAUC was introduced to tackle the early enrichment problem by computing the percentage of the ideal area that lies under the semilog enrichment curve. Formally, we define LogAUCλ, where the log area computations run from λ to 1.0, and in this paper we refer to LogAUC0.001 as simply LogAUCλ.
DrawSth is based on the Python environment, so you should download the package and open the file drawsth.py in your Spyder or Python3. Click on button run and then a graphical user interface will appear. Firstly, select the data file.