def different_window_sizes(W_PRR, W_HISTORY):
print(f'*** PRR={W_PRR}, HISTORY={W_HISTORY} ***')
dataset = load_rutgers()
print('Rutgers loaded ...')
dataset = CustomInterpolation(source='rssi',
strategy='constant',
constant=0).fit_transform(dataset)
print('Interpolation applied ...')
dataset = SyntheticFeatures(source='rssi',
window_size=W_HISTORY).fit_transform(dataset)
print('Synthetic features created ...')
dataset = PRR(source='received', window_size=W_PRR, ahead=1,
target='prr').fit_transform(dataset)
print('PRR created ...')
dataset = CustomMerger().fit_transform(dataset)
print('Datasets merged ...')
dataset['class'] = dataset['prr'].apply(prr_to_label)
print('Classification applied ...')
dataset = dataset.dropna()
X, y = dataset[features].copy(), dataset['class'].ravel()
y_pred = model_selection.cross_val_predict(pipe, X, y, cv=cv, n_jobs=-1)
return y, y_pred
def guassian(pipe):
print('*** Gaussian interpolation ***')
dataset = []
for df in load_rutgers():
df.loc[df['received'] == 0, 'rssi'] = np.nan
dataset.append(df)
print('Rutgers loaded ...')
dataset = CustomInterpolation(source='rssi', strategy='gaussian').fit_transform(dataset)
print('Interpolation applied ...')
dataset = SyntheticFeatures(source='rssi', window_size=10).fit_transform(dataset)
print('Synthetic features created ...')
dataset = PRR(source='received', window_size=10, ahead=1, target='prr').fit_transform(dataset)
print('PRR created ...')
dataset = CustomMerger().fit_transform(dataset)
print('Datasets merged ...')
dataset['class'] = dataset['prr'].apply(prr_to_label)
print('Classification applied ...')
dataset = dataset.dropna()
X, y = dataset[features].copy(), dataset['class'].ravel()
y_pred = model_selection.cross_val_predict(pipe, X, y, cv=cv, n_jobs=-1)
return y, y_pred
def prepare_data():
dataset = load_rutgers()
print('Rutgers loaded ...')
dataset = CustomInterpolation(source='rssi', strategy='constant', constant=0).fit_transform(dataset)
print('Interpolation applied ...')
dataset = SyntheticFeatures(source='rssi', window_size=10).fit_transform(dataset)
print('Synthetic features created ...')
dataset = PRR(source='received', window_size=10, ahead=1, target='prr').fit_transform(dataset)
print('PRR created ...')
for df in dataset:
df['class_overall'] = prr_to_label( df['received'].astype(bool).sum() / 300. )
print('Created special feature `class_overall`')
dataset = CustomMerger().fit_transform(dataset)
print('Datasets merged ...')
dataset['class'] = dataset['prr'].apply(prr_to_label)
print('Classification applied ...')
#dataset = dataset.dropna()
dataset = dataset.sample(frac=1, random_state=SEED)
return dataset
def prepare_data():
dataset = load_rutgers()
print('Rutgers loaded ...')
dataset = CustomInterpolation(source='rssi',
strategy='constant',
constant=0).fit_transform(dataset)
print('Interpolation applied ...')
dataset = SyntheticFeatures(source='rssi',
window_size=20).fit_transform(dataset)
print('Synthetic features created ...')
dataset = PRR(source='received', window_size=20, ahead=1,
target='prr').fit_transform(dataset)
print('PRR created ...')
dataset = CustomMerger().fit_transform(dataset)
print('Datasets merged ...')
dataset['class'] = dataset['prr'].apply(prr_to_label)
print('Classification applied ...')
dataset = dataset.dropna()
dataset = dataset.drop(['noise', 'src', 'dst', 'received', 'prr'], axis=1)
print('Drop useless features, drop lines with NaN')
return dataset
def prepare_data():
dataset = load_rutgers()
print('Rutgers loaded ...')
dataset = CustomInterpolation(source='rssi', strategy='constant', constant=0).fit_transform(dataset)
print('Interpolation applied ...')
dataset = SyntheticFeatures(source='rssi', window_size=10).fit_transform(dataset)
print('Synthetic features created ...')
dataset = PRR(source='received', window_size=10, ahead=1, target='prr').fit_transform(dataset)
print('PRR created ...')
print('Apply discrete derivation (backward difference)')
for i in range(len(dataset)):
dataset[i]['drssi'] = dataset[i]['rssi'].diff()
dataset = CustomMerger().fit_transform(dataset)
print('Datasets merged ...')
dataset['class'] = dataset['prr'].apply(prr_to_label)
print('Classification applied ...')
dataset = dataset.dropna()
X, y = dataset[features].copy(), dataset['class'].ravel()
return X, y
def prepare_data():
dataset = load_rutgers()
print('Rutgers loaded ...')
dataset = CustomInterpolation(source='rssi',
strategy='constant',
constant=0).fit_transform(dataset)
print('Interpolation applied ...')
dataset = SyntheticFeatures(source='rssi',
window_size=10).fit_transform(dataset)
print('Synthetic features created ...')
dataset = PRR(source='received', window_size=10, ahead=1,
target='prr').fit_transform(dataset)
print('PRR created ...')
print('Apply discrete derivation (backward difference)')
for i in range(len(dataset)):
dataset[i]['drssi'] = dataset[i]['rssi'].diff()
dataset = CustomMerger().fit_transform(dataset)
print('Datasets merged ...')
dataset['class'] = dataset['prr'].apply(prr_to_label)
print('Classification applied ...')
dataset = dataset.dropna()
dataset = dataset.drop(['noise', 'src', 'dst', 'received', 'prr'], axis=1)
print('Drop useless features, drop lines with NaN')
dataset = poly_features(dataset,
include=['rssi', 'rssi_avg', 'rssi_std'],
degree=4,
include_bias=True)
print('Polynomials applied ...')
# Special synthetic features
dataset['rssi^-1'] = 1. / dataset['rssi']
dataset['rssi^-2'] = 1. / dataset['rssi^2']
dataset['rssi^-3'] = 1. / dataset['rssi^3']
dataset['rssi^-4'] = 1. / dataset['rssi^4']
dataset['rssi_avg^-1'] = 1. / dataset['rssi_avg']
dataset['rssi_avg^-2'] = 1. / dataset['rssi_avg^2']
dataset['rssi_avg^-3'] = 1. / dataset['rssi_avg^3']
dataset['rssi_avg^-4'] = 1. / dataset['rssi_avg^4']
dataset['rssi_std^-1'] = 1. / dataset['rssi_std']
dataset['rssi_std^-2'] = 1. / dataset['rssi_std^2']
dataset['rssi_std^-3'] = 1. / dataset['rssi_std^3']
dataset['rssi_std^-4'] = 1. / dataset['rssi_std^4']
return dataset
def without(pipe):
print('*** Without interpolation ***')
#dataset =load_rutgers(
dataset = []
for df in load_rutgers():
df.loc[df['received'] == 0, 'rssi'] = np.nan
df.dropna()
dataset.append(df)
print('Rutgers loaded ...')
#dataset = CustomInterpolation(on_column='rssi', strategy='gaussian').fit_transform(dataset)
#print('Interpolation applied ...')
dataset = SyntheticFeatures(source='rssi', window_size=10).fit_transform(dataset)
print('Synthetic features created ...')
dataset = PRR(source='received', window_size=10, ahead=1, target='prr').fit_transform(dataset)
print('PRR created ...')
print('Apply discrete derivation (backward difference)')
for i in range(len(dataset)):
dataset[i]['drssi'] = dataset[i]['rssi'].diff()
dataset = CustomMerger().fit_transform(dataset)
print('Datasets merged ...')
dataset['class'] = dataset['prr'].apply(prr_to_label)
print('Classification applied ...')
dataset = dataset.dropna()
X, y = dataset[features].copy(), dataset['class'].ravel()
y_pred = model_selection.cross_val_predict(pipe, X, y, cv=cv, n_jobs=-1)
return y, y_pred