before_event_mask = dilated_df['time_for_the_event'] < 30
dilated_df = dilated_df.loc[before_event_mask]
# cluster based on time
# cluster_based_on_time(dilated_df['time_for_the_event'], k=2)
betas_values = {}
betas = [13] #np.logspace(-3, 3, 50)
censored_time = {}
removed_rows = {}
for beta in betas:
print('beta = ' + str(beta))
K = calculate_distance(
dilated_df[list(range(20))], {
subject_id: subject_data[list(range(20))]
for subject_id, subject_data in censored_data.items()
}, beta)
K_t_time = K.transpose().join(dilated_df['time_for_the_event'])
def multiply_by_time_for_the_event(col):
return col.apply(lambda x: x * col['time_for_the_event'])
K_t_time_multiplied_time_for_the_event = K_t_time.apply(
multiply_by_time_for_the_event, axis=1)
denominator = K_t_time.sum()
nominator = K_t_time_multiplied_time_for_the_event.sum()
censored_time[beta] = nominator / denominator
censored_data_with_time = OtuMf.mapping_file.loc[
for (subject_id, class_predicted), subject_data in zip(
class_prediction.items(), inputs.values()):
inputs_per_class[class_predicted].update(
{subject_id: subject_data})
K = {}
K_t_time = {}
K_t_time_multiplied_time_for_the_event = {}
censored_data_with_time = pd.DataFrame()
for class_number in range(number_of_classes):
a = dilated_df.loc[dilated_df_clusterd['cluster_number'] ==
class_number]
K[class_number], _ = calculate_distance(
a[list(range(20))],
inputs_per_class[class_number],
beta,
visualize=False)
K_t_time[class_number] = K[class_number].transpose().join(
dilated_df['time_for_the_event'])
K_t_time_multiplied_time_for_the_event[
class_number] = K_t_time[class_number].apply(
multiply_by_time_for_the_event, axis=1)
denominator = K_t_time[class_number].sum()
nominator = K_t_time_multiplied_time_for_the_event[
class_number].sum()
tmp = nominator / denominator
censored_data_with_time_per_class = OtuMf.mapping_file.loc[
tmp.index[:-1].tolist()]
censored_data_with_time_per_class['time_for_the_event'] = tmp
censored_data_with_time = censored_data_with_time.append(
betas_values = {}
beta = 9
censored_time = {}
removed_rows = {}
print('beta = ' + str(beta))
def multiply_by_time_for_the_event(col):
return col.apply(lambda x: x * col['time_for_the_event'])
inputs = {subject_id: subject_data[list(range(n_components))] for subject_id, subject_data in
censored_data.items()}
K, _ = calculate_distance(dilated_df[list(range(n_components))], inputs, beta, visualize=False)
K_t_time = K.transpose().join(dilated_df['time_for_the_event'])
K_t_time_multiplied_time_for_the_event = K_t_time.apply(multiply_by_time_for_the_event, axis=1)
denominator = K_t_time.sum()
nominator = K_t_time_multiplied_time_for_the_event.sum()
censored_time[beta] = nominator / denominator
censored_data_with_time = OtuMf.mapping_file.loc[censored_time[beta].index[:-1].tolist()]
censored_data_with_time['time_for_the_event'] = censored_time[beta]
censored_data_with_time = censored_data_with_time.join(otu_after_pca_wo_taxonomy)
number_of_rows_before_removal = censored_data_with_time.shape[0]
# remove subects with no data in mocrobiome
censored_data_with_time = censored_data_with_time.loc[censored_data_with_time[0].notnull()]