def prepare_training(list_of_training_patients):
# creates a giant hidden sequence array and train_df from all training patients
# create array of hidden state sequence of traning dataset
training_class_array = []
# create df of observation sequence of traning dataset
train_df = pd.DataFrame()
for j in range(list_of_training_patients.shape[0]):
path = data_path + str(list_of_training_patients['file_name'][j])
if os.path.isfile(path) == True:
data = dp.data_import(path)
binary_features = [
"Gain", "Bradycardia", "LegMovement", "CentralApnea",
"Arousal", "Hypopnea", "RelativeDesaturation", "Snore",
"ObstructiveApnea", "MixedApnea", "LongRR", "Tachycardia"
]
for feature in binary_features:
if feature in data.columns:
data = data.drop(feature, axis=1)
df1 = data.pop('hypnogram_User')
data['hypnogram_User'] = df1
data_columns, hidden_sequence, observation_sequence, train, test = preprocess_data(
data=data)
training_class_array.append(hidden_sequence)
train_df = train_df.append(train)
else:
print('File not found.')
pass
feature_names = data.drop(['hypnogram_User', 'hypnogram_Machine'],
axis=1).columns.values.tolist()
del data, observation_sequence, test
return train_df, training_class_array, feature_names,
dist, state_names = hmm_dist.gauss_kernel_dist(feature_names)
print('Observation probabilities prepared')
# * initiate HMM *
model = pg.NaiveBayes(dist)
# * test the model
list_of_testing_patients = list_of_testing_patients.reset_index()
for k in range(list_of_testing_patients.shape[0]):
path = data_path + str(list_of_testing_patients['file_name'][k])
if os.path.isfile(path) == True:
patient_data = dp.data_import(path)
binary_features = ["Gain", "Bradycardia", "LegMovement", "CentralApnea", "Arousal", "Hypopnea",
"RelativeDesaturation", "Snore", "ObstructiveApnea", "MixedApnea", "LongRR", "Tachycardia"]
for feature in binary_features:
if feature in patient_data.columns:
patient_data = patient_data.drop(feature, axis=1)
df1 = patient_data.pop('hypnogram_User')
patient_data['hypnogram_User'] = df1
n_features = patient_data.shape[1] - 2
data_columns, hidden_sequence, observation_sequence, train1, test = preprocess_data(data=patient_data)
test_observation_sequence = train1.iloc[:, 0:n_features].values.tolist()
path = model.predict(test_observation_sequence)
conf_hmm = metrics.confusion_matrix(hidden_sequence, [state_names[id] for id in path], states)
"experimenty/list_of_patients_with_attributes.csv")
#LEARN MODEL ON FIRST TRAINING SET PATIENT
score = []
print("zacal")
for i in range(0, len(list_of_patients['file_name'])):
list_of_testing_patients = list_of_patients.iloc[[i]]
list_of_testing_patients = list_of_testing_patients.reset_index()
list_of_training_patients = list_of_patients.drop([i], axis=0)
list_of_training_patients = list_of_training_patients.reset_index()
print(".")
# PREPROCESS TESTING FILE
testing_patient_path = "Data/" + str(
list_of_testing_patients['file_name'][0])
testing_patient_data = dp.data_import(testing_patient_path)
binary_features = [
"Gain", "Bradycardia", "LegMovement", "CentralApnea", "Arousal",
"Hypopnea", "RelativeDesaturation", "Snore", "ObstructiveApnea",
"MixedApnea", "LongRR", "Tachycardia"
]
for feature in binary_features:
if feature in testing_patient_data.columns:
testing_patient_data = testing_patient_data.drop(feature, axis=1)
df1 = testing_patient_data.pop('hypnogram_User')
testing_patient_data['hypnogram_User'] = df1
testing_patient_data = testing_patient_data.drop(['hypnogram_Machine'],
axis=1)
test_observation_sequence = testing_patient_data.iloc[:, :-1].values
test_hidden_sequence = testing_patient_data.iloc[:, -1].values
import os
import pandas as pd
import hmm as myhmm
import matplotlib.pyplot as plt
import data_preprocessing as dp
#hypnogram of average person
average_person = dp.data_import(
'/Users/kristina/PycharmProjects/vyskumak/Data/12.10.2016-Z-M-39let.csv')
average_person['time'] = [(0.5 / 60 * i) for i in range(0, 973)]
cleanup_nums = {
"hypnogram_User": {
"Wake": 5,
"REM": 4,
"NonREM1": 3,
"NonREM2": 2,
"NonREM3": 1
}
}
average_person.replace(cleanup_nums, inplace=True)
plt.plot(average_person['time'], average_person['hypnogram_User'])
y = [5, 4, 3, 2, 1]
labels = ["Wake", "REM", "NonREM1", "NonREM2", "NonREM3"]
plt.yticks(y, labels)
plt.xlabel('Measurement length [hours]')
plt.ylabel('Sleep stage')
plt.title('Hypnogram of average person from dataset Hradec Kralove')
plt.show()
#age histogram
plt.hist(males['age'], 10, color='grey')
for file in os.listdir(directory):
try:
filename = os.fsdecode(file)
if filename.endswith(".csv") and forbidden_ps1[
forbidden_ps1['Pacient'].str.contains(
filename[:-4])].empty == True:
path = str(directory)[2:-1] + "/" + str(filename)
y_pred = pd.read_csv(path, delim_whitespace=True, header=0)
preds.extend(y_pred['0'])
path1 = hs_dir + str(filename)
y = dp.data_import(path1)
for i in reversed(range(0, len(y['hypnogram_User']))):
if y['hypnogram_User'][i] == "NotScored":
y = y.drop([i])
hs.extend(y['hypnogram_User'])
if len(y_pred['0']) != len(y['hypnogram_User']):
print(
str(filename) + ' ' + str(len(y_pred['0'])) + ' ' +
str(len(y['hypnogram_User'])))
except:
pass
difs = [0 if preds[i] != hs[i] else 1 for i in range(0, len(hs))]
dif = pd.DataFrame({'difs': difs})