def time_series_sum_of_reoccurring_data_points(x):
"""
:param x: the time series to calculate the feature of
:type x: pandas.Series
:return: the value of this feature
:return type: float
"""
return ts_feature_calculators.sum_of_reoccurring_data_points(x)
def TS_features12(signal):
stand_deviation = ts.standard_deviation(signal)
sum_reoccurring = ts.sum_of_reoccurring_data_points(signal)
sum_r_value = ts.sum_of_reoccurring_values(signal)
sum_v = ts.sum_values(signal)
variance = ts.variance(signal)
variance_larger_than_sd = ts.variance_larger_than_standard_deviation(
signal)
return stand_deviation, sum_reoccurring, sum_r_value, sum_v, variance, variance_larger_than_sd
def time_series_sum_of_reoccurring_data_points(x):
"""
Returns the sum of all data points, that are present in the time series
more than once.
:param x: the time series to calculate the feature of
:type x: pandas.Series
:return: the value of this feature
:return type: float
"""
return ts_feature_calculators.sum_of_reoccurring_data_points(x)
def feature_extraction(cgm_glucose):
n = len(cgm_glucose[0])
chunk_size = n // 4
rows = []
for row in cgm_glucose:
val = []
# Feature set 1 : Mean
for i in range(0, 30, 6):
val.append(fc.mean(row[i:i + 6]))
# Feature set 2: Sum of reoccurring data points
for i in range(0, 30, 6):
val.append(fc.sum_of_reoccurring_data_points(row[i:i + 6]))
# Computing the top 5 fft coefficients
fft_coefficients = fft(row, n=6)[1:]
fft_coefficients_real = [value.real for value in fft_coefficients]
val += fft_coefficients_real
# val.extend(fft_coefficients_real)
# Feature set 4: Polyfit
x = np.linspace(0, 1, len(row))
y = row
val.extend(poly.polyfit(x, y, 3)[:-1])
rows.append(val)
feature_matrix = pd.DataFrame(StandardScaler().fit_transform(rows))
# Transform data to training dimensional space
pca_components = pd.read_csv("pca_components.csv", header=None)
transformedData = np.dot(feature_matrix, np.transpose(pca_components))
# Run all 4 classfication algorithms on test data
classify_test_data(transformedData)
def feature_extraction():
cgm_glucose_levels_no_meal = pd.read_csv("preprocessed_Nomeal.csv")
cgm_glucose_levels_meal = pd.read_csv("preprocessed_mealData.csv")
cgm_glucose = np.concatenate(
(cgm_glucose_levels_meal.values, cgm_glucose_levels_no_meal.values),
axis=0)
n = len(cgm_glucose[0])
chunk_size = n // 4
rows = []
for row in cgm_glucose:
val = []
# Feature set 1 : Windowed Mean
for i in range(0, 30, 6):
val.append(fc.mean(row[i:i + 6]))
# Feature set 2: Windowed Variance
# for i in range(0,30,6):
# val.append(fc.variance(row[i:i+6]))
for i in range(0, 30, 6):
val.append(fc.sum_of_reoccurring_data_points(row[i:i + 6]))
# Computing the top 5 fft coefficients
fft_coefficients = fft(row, n=6)[1:]
fft_coefficients_real = [value.real for value in fft_coefficients]
val += fft_coefficients_real
#val.append(np.sqrt(np.mean(row[24:]**2)))
# Feature set 4: Polyfit
x = np.linspace(0, 1, len(row))
y = row
val.extend(poly.polyfit(x, y, 3)[:-1])
rows.append(val)
# for i in range(0, 30, 6):
# val.append(fc.change_quantiles(row[i:i + 6],))
feature_matrix = pd.DataFrame(StandardScaler().fit_transform(rows))
labels = [1] * len(cgm_glucose_levels_meal)
label_no = [0] * len(cgm_glucose_levels_no_meal)
labels.extend(label_no)
labels = np.array(labels)
feature_matrix_meal = feature_matrix.iloc[:len(cgm_glucose_levels_meal
), :].values
feature_matrix_no_meal = feature_matrix.iloc[
len(cgm_glucose_levels_meal):, :].values
pca = PCA()
pca.fit(feature_matrix_meal)
# print("PCA explained variance: ")
# print(pca.explained_variance_ratio_)
# Saving new dim-space
pd.DataFrame(pca.components_[:5]).to_csv("pca_components.csv",
header=None,
index=None)
transformedData = np.dot(feature_matrix_meal,
np.transpose(pca.components_[:5]))
transformedData_no_meal = np.dot(feature_matrix_no_meal,
np.transpose(pca.components_[:5]))
transformedData = np.concatenate(
(transformedData, transformedData_no_meal))
return transformedData, labels
