tsmoothie

1.0.5 verified Fri May 01 auth: no python

A Python library for timeseries smoothing and outlier detection in a vectorized way. It provides multiple smoothing techniques (e.g., Convolutional, Exponential, Lowess, Kalman, Spectral) and outlier detection methods. Current version is 1.0.5. Release cadence is intermittent.

pip install tsmoothie

Common errors

error ImportError: cannot import name 'ConvolutionalSmoother' from 'tsmoothie' ↓

cause The import path changed in v1.0.2; you need to import from submodule.

fix

Use 'from tsmoothie.smoother import ConvolutionalSmoother'

error ValueError: Data must be 2-dimensional ↓

cause Passed a 1D array instead of 2D.

fix

Reshape data: data.reshape(1, -1) for a single series.

error TypeError: smooth() got an unexpected keyword argument 'window' ↓

cause The smoother's __init__ may accept window parameter, but some like LowessSmoother use 'smooth_fraction'.

fix

Check the specific smoother's documentation for correct parameters.

error ModuleNotFoundError: No module named 'tsmoothie' ↓

cause Library not installed.

fix

Run 'pip install tsmoothie'

Warnings

gotcha Data must be 2D array-like of shape (n_series, n_points). Even a single series must be shaped (1, n_points) or a list of lists. ↓

fix If you have a 1D array x, use x.reshape(1, -1) or [x] before passing to smoother.

gotcha Smoothers expect data in float dtype. Integer arrays may cause incorrect results or errors. ↓

fix Convert data to float: data.astype(float).

breaking In v1.0.2, the code was restructured: many imports moved to submodules. Previously some classes might have been importable directly from tsmoothie. ↓

fix Use from tsmoothie.smoother import ... instead of from tsmoothie import ...

deprecated The method 'get_intervals' parameter 'prediction_interval' is still valid, but future versions may change the API. No official deprecation yet. ↓

fix Stay updated with CHANGELOG.

Imports

ConvolutionalSmoother
```
from tsmoothie.smoother import ConvolutionalSmoother
```
All smoothers are in tsmoothie.smoother module

LowessSmoother

from tsmoothie.smoother import LowessSmoother

ExponentialSmoother

from tsmoothie.smoother import ExponentialSmoother

SpectralSmoother

from tsmoothie.smoother import SpectralSmoother

Added in v1.0.1

DecomposeSmoother

from tsmoothie.smoother import DecomposeSmoother

Added in v0.2.0

KalmanSmoother

from tsmoothie.smoother import KalmanSmoother

WindowWrapper
```
from tsmoothie.utils import WindowWrapper
```
Utility for window-based operations

load_dataset

from tsmoothie.datasets import load_dataset

Utility to load example datasets

Quickstart

Basic smoothing with LowessSmoother. Tsmoothie operates on 2D arrays (time series in rows). Ensure data is float and shape (n_series, n_points).

import numpy as np
from tsmoothie.smoother import LowessSmoother

# Generate noisy data
np.random.seed(42)
data = np.cumsum(np.random.randn(100)) + 10

# Set up smoother
smoother = LowessSmoother(smooth_fraction=0.1, iterations=1)
smoother.smooth(data)

# Access smoothed and intervals
smoothed = smoother.smooth_data[0]
low, up = smoother.get_intervals('prediction_interval')

print('Smoothed:', smoothed[:5])