y_train = y_train.astype('float32')
y_test = y_test.astype('float32')
print('Pre-training...')
model_filename = \
os.path.join(output_dir, 'models',
'%s_%s_exp%d_mf%d_msl%d_km%d_test.pkl'
% (init_survival_estimator_name, dataset,
experiment_idx, max_features,
min_samples_leaf, use_km))
if not os.path.isfile(model_filename):
surv_model = \
RandomSurvivalForest(n_estimators=100,
max_features=max_features,
max_depth=None,
oob_score=False,
feature_importance=False,
min_samples_leaf=min_samples_leaf,
random_state=init_random_seed,
n_jobs=n_jobs)
surv_model.fit(X_train, y_train)
surv_model.save(model_filename)
else:
surv_model = RandomSurvivalForest.load(model_filename)
print('*** Extracting proximity matrix...')
prox_filename = model_filename[:-4] + '_prox_matrix.txt'
if not os.path.isfile(prox_filename):
leaf_ids = surv_model.predict_leaf_ids(X_train)
n = len(X_train)
prox_matrix = np.eye(n)
tic = time.time()
model_filename = \
os.path.join(output_dir, 'models',
'%s_%s_exp%d_%s_mf%d_msl%d_fold%d.pkl'
% (survival_estimator_name, dataset,
experiment_idx, val_string,
max_features, min_samples_leaf,
fold_idx))
time_elapsed_filename = model_filename[:-4] + '_time.txt'
if not os.path.isfile(model_filename):
surv_model = \
RandomSurvivalForest(
n_estimators=100,
max_features=max_features,
max_depth=None,
oob_score=False,
feature_importance=False,
min_samples_leaf=min_samples_leaf,
random_state=method_random_seed,
n_jobs=n_jobs)
surv_model.fit(fold_X_train, fold_y_train)
elapsed = time.time() - tic
print('Time elapsed: %f second(s)' % elapsed)
np.savetxt(time_elapsed_filename,
np.array(elapsed).reshape(1, -1))
surv_model.save(model_filename)
else:
surv_model = RandomSurvivalForest.load(model_filename)
elapsed = float(np.loadtxt(time_elapsed_filename))
print('Time elapsed (from previous fitting): ' +
'%f second(s)' % elapsed)
arg_min_IPEC_scores = [None for idx in range(num_IPEC_horizons)]
for n_estimators, max_depth in hyperparams:
cindex_scores = []
IPEC_scores = [[] for idx in range(num_IPEC_horizons)]
for train_idx, val_idx in kf.split(X_train):
fold_X_train = X_train[train_idx]
fold_y_train = y_train[train_idx]
fold_X_val = X_train[val_idx]
fold_y_val = y_train[val_idx]
surv_model = RandomSurvivalForest(n_estimators=n_estimators,
max_depth=max_depth,
oob_score=False,
feature_importance=False,
min_samples_leaf=3,
random_state=rng,
n_jobs=-1)
surv_model.fit(fold_X_train, fold_y_train)
sorted_fold_y_val = np.sort(np.unique(fold_y_val[:, 0]))
if sorted_fold_y_val[0] != 0:
mesh_points = np.concatenate(([0.], sorted_fold_y_val))
else:
mesh_points = sorted_fold_y_val
surv = \
surv_model.predict_surv(fold_X_val, mesh_points,
presorted_times=True)
# -------------------------------------------------------------
sys.exit()
print('[Dataset: %s, experiment: %d]' % (dataset, experiment_idx))
print()
X_train, y_train, X_test, y_test, feature_names, \
compute_features_and_transformer, transform_features = \
load_dataset(dataset, experiment_idx)
print('Testing...', flush=True)
model_filename = \
os.path.join(output_dir, 'models',
'%s_%s_exp%d_%s_mf%d_msl%d_test.pkl'
% (survival_estimator_name, dataset, experiment_idx,
val_string, max_features, min_samples_leaf))
surv_model = RandomSurvivalForest.load(model_filename)
surv_model.n_jobs = n_jobs
sorted_unique_y_train = np.unique(y_train[:, 0])
surv = surv_model.predict_surv(X_test,
sorted_unique_y_train,
presorted_times=True)
print()
print('[Test data statistics]')
sorted_y_test_times = np.sort(y_test[:, 0])
print('Quartiles:')
print('- Min observed time:', np.min(y_test[:, 0]))
print('- Q1 observed time:',
sorted_y_test_times[int(0.25 * len(sorted_y_test_times))])
print('- Median observed time:', np.median(y_test[:, 0]))
print('- Q3 observed time:',
print('Pre-training...')
tic = time.time()
model_filename = \
os.path.join(output_dir, 'models',
'%s_%s_exp%d_%s_mf%d_msl%d_test.pkl'
% (init_survival_estimator_name, dataset,
experiment_idx, init_val_string, max_features,
min_samples_leaf))
time_elapsed_filename = model_filename[:-4] + '_time.txt'
if not os.path.isfile(model_filename):
surv_model = \
RandomSurvivalForest(n_estimators=100,
max_features=max_features,
max_depth=None,
oob_score=False,
feature_importance=False,
min_samples_leaf=min_samples_leaf,
random_state=init_random_seed,
n_jobs=n_jobs)
surv_model.fit(X_train, y_train)
elapsed = time.time() - tic
print('Time elapsed: %f second(s)' % elapsed)
np.savetxt(time_elapsed_filename, np.array(elapsed).reshape(1, -1))
surv_model.save(model_filename)
else:
surv_model = RandomSurvivalForest.load(model_filename)
elapsed = float(np.loadtxt(time_elapsed_filename))
print('Time elapsed (from previous fitting): %f second(s)' %
elapsed)