sktime - 时间序列机器学习统一接口

一、关于 sktime

1、项目概览

sktime 是一个 Python 时间序列分析库，为多种时间序列学习任务提供统一接口。当前支持的功能包括：

• 时间序列预测
• 时间序列分类
• 时间序列聚类
• 异常/变化点检测
• 其他相关任务

该库提供时间序列算法和兼容[scikit-learn]的工具，用于构建、调优和验证时间序列模型。

2、相关链接资源

• Github：https://github.com/sktime/sktime
• 官网：https://www.sktime.net
• 官方文档：https://www.sktime.net/en/stable/users.html
• 教程：https://www.sktime.net/en/stable/examples.html
• Binder Notebooks：https://mybinder.org/v2/gh/sktime/sktime/main?filepath=examples
• YouTube 教程：https://www.youtube.com/playlist?list=PLKs3UgGjlWHqNzu0LEOeLKvnjvvest2d0
• 社区支持：
  • Discord：https://discord.com/invite/54ACzaFsn7
  • Stack Overflow：https://stackoverflow.com/questions/tagged/sktime
  • LinkedIn：https://www.linkedin.com/company/scikit-time/
• License：BSD 3-Clause
• 版本状态：
  • PyPI：https://pypi.org/project/sktime/
  • Conda：https://anaconda.org/conda-forge/sktime
• 引用：Zenodo

二、功能特性

1、核心模块

2、生态系统集成

与以下库实现互操作：

• scikit-learn
• statsmodels
• tsfresh
• PyOD
• fbprophet

三、安装配置

1、pip安装

# 基础安装
pip install sktime

# 完整功能安装
pip install sktime[all_extras]

# 按需安装特定模块
pip install sktime[forecasting]  # 仅预测相关依赖
pip install sktime[forecasting,transformations]  # 预测+转换

2、conda安装

conda install -c conda-forge sktime

# 完整功能安装
conda install -c conda-forge sktime-all-extras

四、快速入门

1、时间序列预测示例

from sktime.datasets import load_airline
from sktime.forecasting.base import ForecastingHorizon
from sktime.forecasting.theta import ThetaForecaster
from sktime.split import temporal_train_test_split
from sktime.performance_metrics.forecasting import mean_absolute_percentage_error

y = load_airline()
y_train, y_test = temporal_train_test_split(y)
fh = ForecastingHorizon(y_test.index, is_relative=False)
forecaster = ThetaForecaster(sp=12)  # monthly seasonal periodicity
forecaster.fit(y_train)
y_pred = forecaster.predict(fh)
mean_absolute_percentage_error(y_test, y_pred)
>>> 0.08661467738190656

2、时间序列分类示例

from sktime.classification.interval_based import TimeSeriesForestClassifier
from sktime.datasets import load_arrow_head
from sklearn.model_selection import train_test_split
from sklearn.metrics import accuracy_score

X, y = load_arrow_head()
X_train, X_test, y_train, y_test = train_test_split(X, y)
classifier = TimeSeriesForestClassifier()
classifier.fit(X_train, y_train)
y_pred = classifier.predict(X_test)
accuracy_score(y_test, y_pred)
>>> 0.8679245283018868

五、社区参与

• 贡献指南：https://www.sktime.net/en/latest/get_involved/contributing.html
• 开发者文档：https://www.sktime.net/en/latest/developer_guide.html
• 社区会议：每周五13:00 UTC（Discord）
• 捐赠支持：https://opencollective.com/sktime

伊织 xAI 2025-05-27（二）