def test(device, use_features, clf_name, with_feature_selection,
activities_to_use, dataset_path):
print(device, use_features, clf_name, str(with_feature_selection),
dataset_path)
file_name = device
if with_feature_selection:
file_name += '_selected'
df = pd.read_pickle(dataset_path + file_name + '.p')
df_labels = pd.read_pickle(dataset_path + device + '_labels.p')
## filter activities
df_labels = df_labels.loc[df_labels.label.isin(activities_to_use)]
df = df.loc[df.index.isin(df_labels.index)]
## filter features
df = df.filter(regex=(use_features))
if len(df.columns) == 0:
print('No features found for')
return None
X_train, y_train, X_test, y_test = split_one_df(df, df_labels, 0.7)
try:
y_pred = classify(X_train, y_train, X_test, clf_name)
except ValueError as ex:
print(ex)
return None
r = log_of_classification_results(y_test, y_pred)
return r
def test(source_device, target_device, dataset_path, use_features,
use_activities, with_feature_selection, clf_name):
source_file_name = source_device + '_selected' if with_feature_selection else source_device
## read features
df_source = pd.read_pickle(dataset_path + source_file_name + '.p')
df_target = pd.read_pickle(dataset_path + target_device + '.p')
## read labels
df_source_labels = pd.read_pickle(dataset_path + source_device +
'_labels.p')
df_target_labels = pd.read_pickle(dataset_path + target_device +
'_labels.p')
## filter activities
df_source, df_source_labels, df_target, df_target_labels = \
filter_by_activities(df_source, df_source_labels, df_target,
df_target_labels, use_activities)
if df_source is None:
return None
## split activities into 70/30
ratio = 0.7
window_ids = df_source_labels['window_id'].tolist(
) + df_target_labels['window_id'].tolist()
window_ids = np.unique(window_ids)
np.random.shuffle(window_ids)
split = int(round(ratio * len(window_ids)))
training_i = window_ids[:split]
testing_i = window_ids[split:]
df_source_labels = df_source_labels.loc[df_source_labels.window_id.isin(
training_i)]
df_target_labels = df_target_labels.loc[df_target_labels.window_id.isin(
testing_i)]
df_source = df_source.loc[df_source.index.isin(df_source_labels.index)]
df_target = df_target.loc[df_target.index.isin(df_target_labels.index)]
df_source, df_target = filter_by_features(df_source, df_target,
use_features)
if df_source is None:
return None
y_source = df_source_labels['label']
y_target = df_target_labels['label']
X_source = X_sort(df_source)
X_target = X_sort(df_target)
try:
y_target_pred = classify(X_source, y_source, X_target, clf_name)
except ValueError as ex:
print(ex)
return None
r = log_of_classification_results(y_target, y_target_pred)
return r
def test(sources, target, use_features, clf_name):
df_sources = [
pd.read_pickle(dataset_path + source + '.p') for source in sources
]
df_target = pd.read_pickle(dataset_path + target + '.p')
df_labels = pd.DataFrame.from_csv(dataset_path + 'activity_labels.csv')
df_sources = [df.filter(regex=(use_features)) for df in df_sources]
df_target = df_target.filter(regex=(use_features))
X_train, y_train, X_test, y_test = \
split_with_multiple_sources(df_sources, df_target, df_labels, 0.7)
y_pred = classify(X_train, y_train, X_test, clf_name)
r = log_of_classification_results(y_test, y_pred)
return r
def test(source, target, use_features, clf_name, target_data_ratio):
df_train = pd.read_pickle(dataset_path + source + '.p')
df_test = pd.read_pickle(dataset_path + target + '.p')
df_labels = pd.DataFrame.from_csv(dataset_path + 'activity_labels.csv')
df_train = df_train.filter(regex=(use_features))
df_test = df_test.filter(regex=(use_features))
X_train, y_train, X_test, y_test = \
split_with_target_training_data(df_source=df_train,
df_target=df_test,
df_labels=df_labels,
ratio=0.7,
target_data_ratio=target_data_ratio)
y_pred = classify(X_train, y_train, X_test, clf_name)
r = log_of_classification_results(y_test, y_pred)
return r
def test(source_device, target_device, source_dataset_path,
target_dataset_path, use_features, use_activities,
with_feature_selection, clf_name):
# print(source_device, target_device, source_dataset_path, target_dataset_path)
source_file_name = source_device + '_selected' if with_feature_selection else source_device
## read features
df_source = pd.read_pickle(source_dataset_path + source_file_name + '.p')
df_target = pd.read_pickle(target_dataset_path + target_device + '.p')
## read labels
df_source_labels = pd.read_pickle(source_dataset_path + source_device +
'_labels.p')
df_target_labels = pd.read_pickle(target_dataset_path + target_device +
'_labels.p')
df_source, df_target = filter_by_features(df_source, df_target,
use_features)
if df_source is None:
return None
## filter activities
df_source, df_source_labels, df_target, df_target_labels = \
filter_by_activities(df_source, df_source_labels, df_target,
df_target_labels, use_activities)
if df_source is None:
return None
y_source = df_source_labels['label']
y_target = df_target_labels['label']
X_source = X_sort(df_source)
X_target = X_sort(df_target)
try:
y_target_pred = classify(X_source, y_source, X_target, clf_name)
except ValueError as ex:
print(ex)
return None
r = log_of_classification_results(y_target, y_target_pred)
return r