base_o.py 文件源码

def get_scores(clf, X_t_train, y_train, X_t_test, y_test):
    clf.fit(X_t_train, y_train)
    y_score = clf.predict_proba(X_t_test)
    app = dict()
    score = fbeta_score(y_test, clf.predict(X_t_test), beta=2, average=None)
    #auc_score = roc_auc_score(y_test, clf.predict(X_t_test), average='samples')
    avg_sample_score = fbeta_score(y_test, clf.predict(X_t_test), beta=2, average='samples')
    prec_score = precision_score(y_test, clf.predict(X_t_test), average='micro')
    rec_score = recall_score(y_test, clf.predict(X_t_test), average='micro')
    avg_prec = average_precision_score(y_test, clf.predict(X_t_test))
    metrics = [score, avg_sample_score, roc_auc_score(y_test, clf.predict_proba(X_t_test))]
    #app['Classwise Scores'] = ([(mlb.classes_[l], score[l]) for l in score.argsort()[::-1]])
    fpr = dict()
    tpr = dict()
    roc_auc = dict()
    for i in range(len(list(enumerate(mlb.classes_)))):
        fpr[i], tpr[i], _ = roc_curve(y_test[:, i], y_score[:, i])
        roc_auc[mlb.classes_[i]] = auc(fpr[i], tpr[i])

    app['F2 Score'] = avg_sample_score
    app['ROC_AUC'] = roc_auc_score(y_test, clf.predict_proba(X_t_test))
    app['Classwise F2 Scores'] = ([(mlb.classes_[l], score[l]) for l in score.argsort()[::-1]])
    app['P_AUPR'] = avg_prec
    app['Precision'] = prec_score
    app['Recall'] = rec_score
    app['ROC_AUC_samples'] = roc_auc
    return app