Performance evaluation plays a crucial role in machine learning. So we may rely on an AUC-ROC Curve for a classification task. The AUC (Area Under The Curve) ROC (Receiver Operating Characteristics) curve is used to assess or depict the performance of the multi-class classification issue. It is one of the most crucial assessment criteria for determining the effectiveness of any classification model. It is also spelled AUROC (Area Under the Receiver Operating Characteristics).
In contrast to micro averaging, which gives equal weight to each sample, macro averaging gives equal weight to each class. Macro and micro will get the same score if there are an equal number of pieces for each class. Recall may be seen as a measure of quantity, whereas Precision can be viewed as a measure of quality. High recall indicates that an algorithm provides the majority of the relevant results. In contrast, higher Precision suggests that an algorithm delivers more relevant results than irrelevant ones (whether or not irrelevant ones are also returned). The harmonic mean of recall and accuracy is what the F1-score is by definition. The following formula combines recall and accuracy into a single number: This formula may also be expressed equivalently. It is important to note that the F1-score accounts for both F.P.s and F.N.s since it considers both accuracy and recall. An examination called a mammogram might determine whether someone has breast cancer. A false positive test would misdiagnose a patient as having breast cancer, while a false negative result would miss a patient who does. An outcome where the model correctly predicted the positive class is considered a true positive. Similar to a genuinely positive, a true negative is a result for which the model accurately foresees the negative class. The precision-recall curve shows the trade-off between accuracy and recall for various thresholds. High accuracy is correlated with a low false positive rate, while big recall is correlated with a low false negative rate. An elevated area under the curve denotes both high recall and High Precision. The trade-off between the true positive rate and the positive predictive value for a predictive model employing various probability thresholds is summarized by precision-recall curves.
The x-axis of a PR curve shows Precision values while the y-axis shows Recall values. TP/(TP + FN) is found on the y-axis of the PR curve, whereas TP/(TP + FP) is located on the x-axis. It's worth noting that Precision is sometimes referred to as "positive predictive value" (PPV).
The Fig. 1 draws the breast cancer prediction ROC for an Ada Boost Classifier. The AUC evaluation criterion for healthy people was 098. alsoAUC for people with tumors was 0.98. Finally, the evaluation criteria of Micro average and Macro average took the value of 0.99.
Brest cancer Prediction Classification Report for an Ada Boost Classifier is shown in Fig. 2. 148 healthy people and 80 people with tumors were considered for test data. Precision, recall, and f1 score evaluation criteria were obtained for healthy and sick people. This algorithm had good accuracy in diagnosing the disease.
The trade-off between the true positive rate and the positive predictive value for a predictive model employing various probability thresholds is summarized by precision-recall curves. The precision-recall curve is created by calculating and displaying the recall against the accuracy for a single classifier at different thresholds. The precision-recall curve for a Brest cancer prediction using an Ada Boost Classifier is shown in Fig. 3. It was 0.98 on average for the accuracy assessment metric.
The Fig. 4 draws the breast cancer prediction ROC for a Random Forest Classifier. The AUC evaluation criterion for healthy people was 099. alsoAUC for people with tumors was 0.99. Finally, the evaluation criteria of Micro average and Macro average took the value of 0.99.
Brest cancer Prediction Classification Report for a Random Forest Classifier is shown in Fig. 5. One hundred forty-eight healthy people and 80 people with tumors were considered for test data. Precision, recall, and f1 score evaluation criteria were obtained for healthy and sick people. This algorithm had the best accuracy in diagnosing the disease. Because it had the best performance in diagnosing people with tumors, in medicine, if a healthy person is diagnosed as sick, the doctors may realize their mistake in the subsequent tests of the person and say that the person is healthy. But if a sick person is mistakenly diagnosed as healthy, there will be no way back, and the sick person will die.
Figure 6 shows the Brest cancer Prediction Precision-Recall Curve for a Random Forest Classifier. The average precision evaluation measure was 0.99. This algorithm had the best accuracy in diagnosing the disease.
Breast cancer prediction ROC for a Logistic Classifier is drawn in the Fig. 7. The AUC evaluation criterion for healthy people was 096. alsoAUC for people with tumors was 0.97. Finally, the evaluation criteria of Micro average and Macro average took the value of 0.97.
Brest cancer Prediction Classification Report for a Logistic Classifier is shown in Fig. 8. 148 healthy people and 80 people with tumors were considered for test data. Precision, recall, and f1 score evaluation criteria were obtained for healthy and sick people. This algorithm had good accuracy in diagnosing the disease.
Figure 9 shows the Brest cancer Prediction Precision-Recall Curve for a Logistic Classifier. The average precision evaluation measure was 0.98.
The Fig. 10 draws the breast cancer prediction ROC for a Gaussian NB Classifier. The AUC evaluation criterion for healthy people was 091. alsoAUC for people with tumors was 0.97. Finally, the evaluation criteria for the Micro average was 0.95, and the Macro average took the value of 0.94.
Brest cancer Prediction Classification Report for an Ada Boost Classifier is shown in Fig. 11. One hundred forty-eight healthy people and 80 people with tumors were considered for test data. Precision, recall, and f1 score evaluation criteria were obtained for healthy and sick people. This algorithm had good accuracy in diagnosing the disease. The worst performance was related to this algorithm. The calculated evaluation criteria prove this claim.
Figure 12 shows the Brest cancer Prediction Precision-Recall Curve for a Gaussian NB Classifier. The average precision evaluation measure was 0.92. This algorithm had the worst accuracy in diagnosing the disease.
The Fig. 13 draws the breast cancer prediction ROC for a Gaussian NB Classifier. The AUC evaluation criterion for healthy people was 098. alsoAUC for people with tumors was 0.98. Finally, the evaluation criteria for the Micro average was 0.98, and the Macro average took the value of 0.98.
Brest cancer Prediction Classification Report for a Gradient Boosting Classifier is shown in Fig. 14. One hundred forty-eight healthy people and 80 people with tumors were considered for test data. Precision, recall, and f1 score evaluation criteria were obtained for healthy and sick people.
Figure 15 shows the Brest cancer Prediction Precision-Recall Curve for a Gradient Boosting Classifier. The average precision evaluation measure was 0.98.