def circuitSNP_metrics(predictions, labels, model_name):
# y_pred = model.predict(x, verbose=0)
y_pred = predictions
y = labels
mets = CH.get_metrics(y, y_pred)
prec, rec, _ = precision_recall_curve(y, y_pred)
auPRC = average_precision_score(y, y_pred)
fpr, tpr, _ = roc_curve(y, y_pred)
auROC = auc(fpr, tpr)
prec_at_10_rec = prec[np.abs(rec - 0.1).argmin()]
auprc = '{:0.4f}'.format(auPRC)
auroc = '{:0.4f}'.format(auROC)
prec_10 = '{:0.4f}'.format(prec_at_10_rec)
rec_50 = '{:0.4f}'.format(mets['rec50'])
return ({
'prec': prec,
'rec': rec,
'auprc': auprc,
'fpr': fpr,
'tpr': tpr,
'auroc': auroc,
'y_pred': y_pred,
'prec_10': prec_10,
'model_name': model_name,
'rec_50': rec_50
})
def get_prc_roc_validation_pwm_model(model, data1, data2, data3,
good_pwms_idx):
# from CNN_Models import cnn_helpers2 as CH
x = [
data1['test_data_X'][:, good_pwms_idx],
data2['test_data_X'][:, good_pwms_idx], data3['test_data_X']
]
y = data1['test_data_Y']
y_pred = model.predict(x, verbose=0)
mets = CH.get_metrics(y, y_pred)
prec, rec, _ = precision_recall_curve(y, y_pred)
auPRC = average_precision_score(y, y_pred)
fpr, tpr, _ = roc_curve(y, y_pred)
auROC = auc(fpr, tpr)
auprc = '{:0.4f}'.format(auPRC)
auroc = '{:0.4f}'.format(auROC)
return ({
'prec': prec,
'rec': rec,
'auprc': auprc,
'fpr': fpr,
'tpr': tpr,
'auroc': auroc,
'y_pred': y_pred
})
def get_prc_roc_test_joint_model(model, x1, x2, x3, y):
y_pred = model.predict([x1, x2, x3], verbose=0)
mets = CH.get_metrics(y, y_pred)
prec, rec, _ = precision_recall_curve(y, y_pred)
auPRC = average_precision_score(y, y_pred)
fpr, tpr, _ = roc_curve(y, y_pred)
auROC = auc(fpr, tpr)
auprc = '{:0.4f}'.format(auPRC)
auroc = '{:0.4f}'.format(auROC)
return ({
'prec': prec,
'rec': rec,
'auprc': auprc,
'fpr': fpr,
'tpr': tpr,
'auroc': auroc,
'y_pred': y_pred
})
def get_prc_roc_validation(model, data):
x = data['test_data_X']
y = data['test_data_Y']
y_pred = model.predict(x, verbose=0)
mets = CH.get_metrics(y, y_pred)
prec, rec, _ = precision_recall_curve(y, y_pred)
auPRC = average_precision_score(y, y_pred)
fpr, tpr, _ = roc_curve(y, y_pred)
auROC = auc(fpr, tpr)
prec_at_10_rec = prec[np.abs(rec - 0.1).argmin()]
auprc = '{:0.4f}'.format(auPRC)
auroc = '{:0.4f}'.format(auROC)
prec_10 = '{:0.4f}'.format(prec_at_10_rec)
return ({
'prec': prec,
'rec': rec,
'auprc': auprc,
'fpr': fpr,
'tpr': tpr,
'auroc': auroc,
'y_pred': y_pred,
'prec_10': prec_10
})
def circuitSNP_test_new_model_metrics(test_data, test_y, model):
x = test_data
y_pred = model.predict(x, verbose=0)
y = test_y
mets = CH.get_metrics(y, y_pred)
prec, rec, _ = precision_recall_curve(y, y_pred)
auPRC = average_precision_score(y, y_pred)
fpr, tpr, _ = roc_curve(y, y_pred)
auROC = auc(fpr, tpr)
prec_at_10_rec = prec[np.abs(rec - 0.1).argmin()]
auprc = '{:0.4f}'.format(auPRC)
auroc = '{:0.4f}'.format(auROC)
prec_10 = '{:0.4f}'.format(prec_at_10_rec)
return ({
'prec': prec,
'rec': rec,
'auprc': auprc,
'fpr': fpr,
'tpr': tpr,
'auroc': auroc,
'y_pred': y_pred,
'prec_10': prec_10
})