def test_derivative_nonlinear_index(self): """Should take into account dt in index when computing derivative""" test_df = pd.DataFrame({"x": [1, 2, 3]}, index=[1, 2, 4]) transform_result = streams.derivative(test_df) self.assertEqual(list(transform_result.columns.values), ["x", "dx_dt"]) self.assertTrue(np.isnan(transform_result.dx_dt[1])) self.assertEqual(transform_result.dx_dt[2], 1.0) self.assertEqual(transform_result.dx_dt[4], 0.5)
def test_derivative(self): """Should compute derivative for default column""" test_df = pd.DataFrame({"x": [1, 2, 3], "y": [4, 5, 6]}) transform_result = streams.derivative(test_df) self.assertEqual(list(transform_result.columns.values), ["x", "y", "dx_dt"]) self.assertTrue(np.isnan(transform_result.dx_dt[0])) self.assertEqual(transform_result.dx_dt[1], 1.0) self.assertEqual(transform_result.dx_dt[2], 1.0)
def test_derivative_alternative_column_name(self): """Should compute derivative of given column name""" test_df = pd.DataFrame({"x": [1, 2, 3], "y": [4, 6, 8]}) transform_result = streams.derivative(test_df, derivative_colname="y") self.assertEqual(list(transform_result.columns.values), ["x", "y", "dy_dt"]) self.assertTrue(np.isnan(transform_result.dy_dt[0])) self.assertEqual(transform_result.dy_dt[1], 2.0) self.assertEqual(transform_result.dy_dt[2], 2.0)
def __total_deviation(self, activity_id): stream_df = load_stream(self.database_driver, activity_id, 'latlng') if stream_df is not None: stream_df = lat_long_to_x_y(stream_df) stream_df = smooth(stream_df, smooth_colname='x') stream_df = smooth(stream_df, smooth_colname='y') stream_df = derivative(stream_df, derivative_colname='x_smooth') stream_df = derivative(stream_df, derivative_colname='y_smooth') stream_df = rolling_similarity(stream_df, cosine_similarity, 'dx_smooth_dt', 'dy_smooth_dt') try: return np.nansum([cosine_to_deviation(cos) for cos in stream_df.cosine_similarity_dx_smooth_dt_dy_smooth_dt]) except ValueError: logging.warning('Failed to compute route deviation for activity %d, returning NaN' % activity_id) return np.nan else: logging.warning('No gps stream available for activity %d, returning NaN' % activity_id) return np.nan