def test_strava_power_model_auto_compute():
# check that the acceleration and the elevation will be auto-computed
power_auto = strava_power_model(activity, cyclist_weight=70)
activity_ele_acc = activity.copy()
activity_ele_acc = gradient_elevation(activity)
activity_ele_acc = acceleration(activity_ele_acc)
power_ele_acc = strava_power_model(activity_ele_acc, cyclist_weight=70)
assert_series_equal(power_auto, power_ele_acc)
# License: BSD 3 clause
###############################################################################
# We can first grab a file and read all data available
from skcycling.datasets import load_fit
from skcycling.io import bikeread
ride = bikeread(load_fit()[0], drop_nan='columns')
###############################################################################
# We can use a physical model to infer the power.
from skcycling.model import strava_power_model
power = strava_power_model(ride, cyclist_weight=72)
###############################################################################
# We can plot the measured and estimated power to observe the difference. We
# can also compute the median absolute error to asses the quality of the
# estimation. To ease the interpretation, we will first resample the data.
import matplotlib.pyplot as plt
from sklearn.metrics import median_absolute_error
resampling_rate = '10S'
measured_power = (ride['power'].resample(resampling_rate)
.mean()
.rename('Measured Power'))
estimated_power = (power.resample(resampling_rate)
.mean()
groups = []
for group_idx, activity in enumerate(data):
groups += [group_idx] * activity.shape[0]
groups = np.array(groups)
weight_user = {'user_1': 86., 'user_2': 72., 'user_3': 61., 'user_5': 72.}
weight_groups = np.zeros_like(groups)
for idx, filename in enumerate(valid_filenames):
mask = groups == idx
weight = [value for key, value in weight_user.items() if key in filename]
weight_groups[mask] = weight[0]
y_pred = np.zeros_like(y.values)
for idx_activity in range(len(valid_filenames)):
mask = groups == idx_activity
y_pred[mask] = strava_power_model(X.iloc[mask, :],
np.unique(weight_groups[mask]),
use_acceleration=True).values
# Store the prediction for visualization
path_results = os.path.join('results', 'mathematical_model')
if not os.path.exists(path_results):
os.makedirs(path_results)
f = os.path.join(path_results, 'y_pred.csv')
pd.Series(y_pred, index=y.index).to_csv(f)
f = os.path.join(path_results, 'y_true.csv')
y.to_csv(f)
np.save(os.path.join(path_results, 'groups.npy'), groups)
f = os.path.join(path_results, 'altitude.csv')
X['elevation'].to_csv(f)
# License: BSD 3 clause
###############################################################################
# We can first grab a file and read all data available
from skcycling.datasets import load_fit
from skcycling.io import bikeread
ride = bikeread(load_fit()[0], drop_nan='columns')
###############################################################################
# We can use a physical model to infer the power.
from skcycling.model import strava_power_model
power = strava_power_model(ride, cyclist_weight=72)
###############################################################################
# We can plot the measured and estimated power to observe the difference. We
# can also compute the median absolute error to asses the quality of the
# estimation. To ease the interpretation, we will first resample the data.
import matplotlib.pyplot as plt
from sklearn.metrics import median_absolute_error
resampling_rate = '10S'
measured_power = (
ride['power'].resample(resampling_rate).mean().rename('Measured Power'))
estimated_power = (
power.resample(resampling_rate).mean().rename('Estimate Power'))
def test_strava_power_model():
# at constant speed the acceleration should not have any influence
power_without_acc = strava_power_model(activity,
cyclist_weight=78,
use_acceleration=False)
power_with_acc = strava_power_model(activity,
cyclist_weight=78,
use_acceleration=True)
assert_allclose(power_without_acc, power_with_acc)
# increase cyclist weight should increase power
power_initial = strava_power_model(activity, cyclist_weight=70)
power_increase_weight = strava_power_model(activity, cyclist_weight=78)
assert_array_less(power_initial, power_increase_weight)
# increase bike weight should increase power
power_initial = strava_power_model(activity,
cyclist_weight=70,
bike_weight=7)
power_increase_weight = strava_power_model(activity,
cyclist_weight=70,
bike_weight=8)
assert_array_less(power_initial, power_increase_weight)
# increase the rolling coefficient should increase power
power_initial = strava_power_model(activity,
cyclist_weight=70,
coef_roll_res=0.0045)
power_increase_cr = strava_power_model(activity,
cyclist_weight=70,
coef_roll_res=0.006)
assert_array_less(power_initial, power_increase_cr)
# increase of the pressure should increase power
power_initial = strava_power_model(activity,
cyclist_weight=70,
pressure=101325)
power_increase_pressure = strava_power_model(activity,
cyclist_weight=70,
pressure=110000)
assert_array_less(power_initial, power_increase_pressure)
# decrease the temperature should increase the power
power_initial = strava_power_model(activity,
cyclist_weight=70,
temperature=15.0)
power_decrease_temperature = strava_power_model(activity,
cyclist_weight=70,
temperature=10.0)
assert_array_less(power_initial, power_decrease_temperature)
# increase the drag coefficient should increase the power
power_initial = strava_power_model(activity,
cyclist_weight=70,
coef_drag=0.5)
power_increase_coef_drag = strava_power_model(activity,
cyclist_weight=70,
coef_drag=1.0)
assert_array_less(power_initial, power_increase_coef_drag)
# increase the rider frontal surface should increase the power
power_initial = strava_power_model(activity,
cyclist_weight=70,
surface_rider=0.32)
power_increase_surface_rider = strava_power_model(activity,
cyclist_weight=70,
surface_rider=0.5)
assert_array_less(power_initial, power_increase_surface_rider)
def test_strava_power_model_error(activity_corrupted, use_acceleration):
with pytest.raises(MissingDataError):
strava_power_model(activity_corrupted,
cyclist_weight=70,
use_acceleration=use_acceleration)