Change Point Generator

This module contains the class for generating synthetic data with different types of change points.

Change Point Generator Class

class source.generator.change_point_generator.ChangePointGenerator(num_segments=3, segment_length=500, change_point_type='sudden_shift', seed=42)

Bases: object

A class to generate time series data with different types of change points.

Parameters:

  • num_segments (int) – Number of segments in the time series data.

  • segment_length (int) – Length of each segment.

  • change_point_type (str) – Type of change point to introduce (‘sudden_shift’, ‘gradual_drift’, ‘periodic_change’).

  • seed (int) – Random seed for reproducibility.

Initializes the ChangePointGenerator with the specified parameters.

Parameters:

  • num_segments (int) – Number of segments in the time series data.

  • segment_length (int) – Length of each segment.

  • change_point_type (str) – Type of change point to introduce (‘sudden_shift’, ‘gradual_drift’, ‘periodic_change’).

  • seed (int) – Random seed for reproducibility.

__init__(num_segments=3, segment_length=500, change_point_type='sudden_shift', seed=42)

Initializes the ChangePointGenerator with the specified parameters.

Parameters:

  • num_segments (int) – Number of segments in the time series data.

  • segment_length (int) – Length of each segment.

  • change_point_type (str) – Type of change point to introduce (‘sudden_shift’, ‘gradual_drift’, ‘periodic_change’).

  • seed (int) – Random seed for reproducibility.

add_gradual_drift(mean_start, mean_end, std_dev, change_point_index)

Add a gradual drift change point to the data.

Parameters:

  • mean_start (float) – Mean value at the start of the drift.

  • mean_end (float) – Mean value at the end of the drift.

  • std_dev (float) – Standard deviation during the drift.

  • change_point_index (int) – Index at which the drift starts.

add_periodic_change(amplitude, period, std_dev, change_point_index)

Add a periodic change point to the data.

Parameters:

  • amplitude (float) – Amplitude of the periodic change.

  • period (float) – Period of the periodic change.

  • std_dev (float) – Standard deviation during the periodic change.

  • change_point_index (int) – Index at which the periodic change starts.

add_sudden_shift(mean_before, mean_after, std_dev_before, std_dev_after, change_point_index)

Add a sudden shift change point to the data.

Parameters:

  • mean_before (float) – Mean value before the change point.

  • mean_after (float) – Mean value after the change point.

  • std_dev_before (float) – Standard deviation before the change point.

  • std_dev_after (float) – Standard deviation after the change point.

  • change_point_index (int) – Index at which the change point occurs.

generate_block_nans(percentage, min_block_size, max_block_size)

Generate data with NaN values in contiguous blocks.

Parameters:

  • percentage (float) – Percentage of the data to be replaced with NaNs (between 0 and 1).

  • min_block_size (int) – Minimum size of each block of NaNs.

  • max_block_size (int) – Maximum size of each block of NaNs.

generate_data()

Generate time series data with different types of change points.

generate_point_nans(percentage)

Generate data with a specified percentage of NaN values.

Parameters:

percentage (float) – Percentage of NaN values to introduce in the data (between 0 and 1).

Returns:

data_with_nans – Data with NaN values introduced.

Return type:

numpy array

get_data()

Returns the generated time series data.

Returns:

data – The generated time series data.

Return type:

numpy array

plot_data()

Plot the generated time series data.

plot_data_with_nans(data_with_nans)

Plot the generated time series data with NaN values.

Parameters:

data_with_nans (numpy array) – Data with NaN values to plot.

Example Usage

Generate Data

from source.generator.change_point_generator import ChangePointGenerator

# Initialize generator
generator = ChangePointGenerator(num_segments=2, segment_length=1000, change_point_type='gradual_drift')

# Generate synthetic data
generator.generate_data()

# Plot generated data
generator.plot_data()
Data Generator

With Point Missingness

# Add point missingness
nan_percentage = 0.25
data_with_nans = generator.generate_point_nans(nan_percentage)

# Plot data with missing values
generator.plot_data_with_nans(data_with_nans)
Point Missingness

With Block Missingness

# Add block missingness
nan_percentage = 0.1
min_block_size=2
max_block_size=50
data_with_nans = generator.generate_block_nans(nan_percentage, min_block_size, max_block_size)

# Plot data with missing values
generator.plot_data_with_nans(data_with_nans)
Block Missingness