tskit is a collection of tools for time series data analysis.

Only univariate time series are supported at the moment.

import pandas as pd

import tskit

ts = (
    tskit.TimeSeries.from_generators(
        ['sine', 'square', 'random_walk'],
        generator_args=[{}, {'period': 1440}, {}],
        weights=[1.0, 0.5, 0.8],
        standardize_idx=[2],
        start=pd.Timestamp('2022-09-27 00:00:00'),
        length=1440,
        freq='1min',
    )
    .smooth('ewma', alpha=0.5)
    .to_shapelet(alpha=5)
    .tile(10)
    .add_noise('gaussian', amplitude=0.2)
)
ts.save('out/uts.csv')

tskit.plot(ts)

• Time series generation
  • Infer the frequency of time series
  • From configs
  • From generators
    • Random walk
    • Sine
    • Cosine
    • Linear
    • Gaussian
    • Uniform
    • Step
    • Exponential
    • Poisson
    • LogNormal
    • Gamma
    • Square
    • Sawtooth
    • Triangle
    • Impulse
    • Perlin
  • From Pandas DataFrames
  • From CSV files
• Add noise to time series
  • Gaussian noise
  • Uniform noise
  • 1-D Perlin noise
• Plot time series
  • Plot univariate time series
• Time series smoothing
  • Moving average (MA)
  • Exponential weighted moving average (EWMA)
  • Median filter
  • Savitzky-Golay filter
• Time series transformation
  • Merge multiple time series
  • Tile time series
  • Shapelet transform
  • Standardize
  • Seasonal-trend decomposition with LOESS (STL)
  • Split time series into segments
  • Interpolate missing values
• Anomaly/outlier injection
  • Point-wise outlier
    • Contextual outlier
    • Global outlier
  • Pattern-wise outlier
    • Shapelet outlier
    • Trend outlier
    • Seasonal outlier
• Support for multivariate time series