def plot_metrics(metrics,
truth_and_predictions,
target_genes,
run_label,
color=None,
filename_prefix=None,
learn_options=None):
if learn_options["metric"] == 'AUC':
best = truth_and_predictions[0] #[np.argmax(cv_scores)]
plt.figure('ROC per gene')
plt.figure('global ROC')
plt.figure('AUC ROC per gene')
all_truth = np.array([])
all_predictions = np.array([])
AUCs = []
AUCs_labels = []
for i, gene in enumerate(target_genes):
if len(best[1][gene]) == 0:
continue
plt.figure('ROC per gene')
plt.subplot(331 + i)
fpr, tpr, _ = sklearn.metrics.roc_curve(best[0][gene],
best[1][gene])
np.savetxt('../results/%s_ROC.txt' % gene,
np.hstack((fpr[:, None], tpr[:, None])))
roc_auc = sklearn.metrics.auc(fpr, tpr)
AUCs.append(roc_auc)
AUCs_labels.append(gene)
plt.plot(fpr, tpr, label=run_label)
plt.title(gene)
h1 = plt.figure('global ROC')
plt.plot(fpr, tpr, color=color, alpha=.2, linewidth=2.)
all_truth = np.hstack((all_truth, best[0][gene]))
all_predictions = np.hstack((all_predictions, best[1][gene]))
plt.legend(loc=0)
plt.figure('AUC ROC per gene')
ax = plt.subplot(111)
rect = ax.bar(range(len(AUCs)), AUCs, width=0.8)
autolabel(ax, rect)
ax.set_ylim((0.5, 1.0))
ax.set_ylabel('AUC ROC')
ax.set_xticks(np.array(range(len(AUCs))) + 0.8 / 2)
ax.set_xticklabels([t for t in AUCs_labels])
fpr, tpr, _ = sklearn.metrics.roc_curve(all_truth, all_predictions)
roc_auc = sklearn.metrics.auc(fpr, tpr)
#print run_label, roc_auc
plt.figure('global ROC')
plt.plot(fpr,
tpr,
label=run_label + " AUC=%.2f" % roc_auc,
color=color,
linewidth=2.)
plt.legend(loc=0)
plt.xlabel('False Positive Rate')
plt.ylabel('True Positive Rate')
#np.savetxt('../results/global_ROC.txt', np.hstack((fpr[:, None], tpr[:, None])))
#np.savetxt('../results/AUCs.txt', np.hstack((np.array([t for t in target_genes])[:, None], np.array(AUCs)[:, None])), fmt='%s')
if filename_prefix != None:
plt.figure('global ROC')
plt.savefig(filename_prefix + 'globalROC.png')
plt.figure('ROC per gene')
plt.savefig(filename_prefix + 'ROC_per_gene.png')
plt.figure('AUC ROC per gene')
plt.savefig(filename_prefix + 'AUCROC_barplot.png')
return roc_auc
else:
plt.figure('NDCG per gene')
ax = plt.subplot(111)
rect = ax.bar(range(len(metrics)), metrics, width=0.8)
autolabel(ax, rect)
ax.set_ylim((0.0, 1.2))
ax.set_ylabel('NDCG')
ax.set_xticks(np.array(range(len(metrics))) + 0.8 / 2)
ax.set_xticklabels([t for t in target_genes])
truth, predictions = truth_and_predictions[0]
all_truth = np.array([])
all_predictions = np.array([])
for i, gene in enumerate(target_genes):
if len(predictions[gene]) == 0:
continue
all_truth = np.hstack((all_truth, truth[gene]))
all_predictions = np.hstack((all_predictions, predictions[gene]))
sorted = all_predictions[np.argsort(all_truth).flatten()[::-1]]
sortedgt = np.sort(all_truth).flatten()[::-1]
NDCG_total = ranking_metrics.ndcg_at_k_custom_n(
sorted, learn_options["NDGC_k"], sortedgt)
if filename_prefix != None:
plt.figure('NDCG per gene')
plt.savefig(filename_prefix + 'NDCG_barplot.png')
return NDCG_total
def plot_metrics(metrics, truth_and_predictions, target_genes, run_label, color=None, filename_prefix=None, learn_options=None):
if learn_options["metric"] == 'AUC':
best = truth_and_predictions[0]#[np.argmax(cv_scores)]
plt.figure('ROC per gene')
plt.figure('global ROC')
plt.figure('AUC ROC per gene')
all_truth = np.array([])
all_predictions = np.array([])
AUCs = []
AUCs_labels = []
for i, gene in enumerate(target_genes):
if len(best[1][gene])==0:
continue
plt.figure('ROC per gene')
plt.subplot(331+i)
fpr, tpr, _ = sklearn.metrics.roc_curve(best[0][gene], best[1][gene])
np.savetxt('../results/%s_ROC.txt' % gene, np.hstack((fpr[:, None], tpr[:, None])))
roc_auc = sklearn.metrics.auc(fpr, tpr)
AUCs.append(roc_auc)
AUCs_labels.append(gene)
plt.plot(fpr, tpr, label=run_label)
plt.title(gene)
h1 = plt.figure('global ROC')
plt.plot(fpr, tpr, color=color, alpha=.2, linewidth=2.)
all_truth = np.hstack((all_truth, best[0][gene]))
all_predictions = np.hstack((all_predictions, best[1][gene]))
plt.legend(loc=0)
plt.figure('AUC ROC per gene')
ax = plt.subplot(111)
rect = ax.bar(range(len(AUCs)), AUCs, width=0.8)
autolabel(ax,rect)
ax.set_ylim((0.5, 1.0))
ax.set_ylabel('AUC ROC')
ax.set_xticks(np.array(range(len(AUCs))) + 0.8 / 2)
ax.set_xticklabels([t for t in AUCs_labels])
fpr, tpr, _ = sklearn.metrics.roc_curve(all_truth, all_predictions)
roc_auc = sklearn.metrics.auc(fpr, tpr)
#print run_label, roc_auc
plt.figure('global ROC')
plt.plot(fpr, tpr, label=run_label + " AUC=%.2f" % roc_auc, color=color, linewidth=2.)
plt.legend(loc=0)
plt.xlabel('False Positive Rate')
plt.ylabel('True Positive Rate')
#np.savetxt('../results/global_ROC.txt', np.hstack((fpr[:, None], tpr[:, None])))
#np.savetxt('../results/AUCs.txt', np.hstack((np.array([t for t in target_genes])[:, None], np.array(AUCs)[:, None])), fmt='%s')
if filename_prefix != None:
plt.figure('global ROC')
plt.savefig(filename_prefix+'globalROC.png')
plt.figure('ROC per gene')
plt.savefig(filename_prefix+'ROC_per_gene.png')
plt.figure('AUC ROC per gene')
plt.savefig(filename_prefix+'AUCROC_barplot.png')
return roc_auc
else:
plt.figure('NDCG per gene')
ax = plt.subplot(111)
rect = ax.bar(range(len(metrics)), metrics, width=0.8)
autolabel(ax,rect)
ax.set_ylim((0.0, 1.2))
ax.set_ylabel('NDCG')
ax.set_xticks(np.array(range(len(metrics))) + 0.8 / 2)
ax.set_xticklabels([t for t in target_genes])
truth, predictions = truth_and_predictions[0]
all_truth = np.array([])
all_predictions = np.array([])
for i, gene in enumerate(target_genes):
if len(predictions[gene])==0:
continue
all_truth = np.hstack((all_truth, truth[gene]))
all_predictions = np.hstack((all_predictions, predictions[gene]))
sorted = all_predictions[np.argsort(all_truth).flatten()[::-1]]
sortedgt = np.sort(all_truth).flatten()[::-1]
NDCG_total = ranking_metrics.ndcg_at_k_custom_n(sorted, learn_options["NDGC_k"], sortedgt)
if filename_prefix != None:
plt.figure('NDCG per gene')
plt.savefig(filename_prefix+'NDCG_barplot.png')
return NDCG_total