Skip to content
Start free
Start free
Data models > Time series

Time Series Data Decomposition

Time series decomposition is a statistical technique that helps with understanding the inherent patterns within time series data, such as identifying the long-term direction (trend), recurring patterns at fixed intervals (seasonality), and randomness (irregular fluctuations):

  • Seasonality: The seasonal component captures regular patterns that repeat over fixed periods, such as daily, monthly, or quarterly fluctuations. Seasonality is influenced by various factors like weather, holidays, and business cycles.
  • Trend: The trend component represents the long-term progression of the series, showing how the data moves up or down over a lengthy period. It's essentially the underlying direction in which the data is heading, excluding shorter-term fluctuations.
  • Residual (Irregular Component): After the trend and seasonal components are removed from the original time series, the remainder is the residual component. It consists of irregular or random fluctuations that cannot be attributed to the trend or seasonal effects. This component can sometimes reveal unusual events or outliers not explained by the trend or seasonality.
By isolating and examining these key components, time series decomposition enhances the accuracy and efficacy of our forecasting models. Time series decomposition is a vital tool for analyzing the evolution of these components over time and for devising strategies to manage each element effectively.
 

Practical Application of Data Decomposition with PyCaret and CrateDB

In the next Jupyter Notebook, we build on our previous example of using PyCaret and CrateDB to showcase time series decomposition. The notebook demonstrates how to use SQLAlchemy to load time series data from CrateDB, preprocess it, and plot the decomposition using PyCaret, as we will explore in the following steps.
 
 

Steps 1 to 6

To follow this tutorial, repeat Steps 1 through 6 from the explanation on Exploratory Data Analysis. Once the environment is set up, proceed to the next step for visualizing time series decomposition.

Step 7. Visualizing Time Series Decomposition

We employ PyCaret to automate the analytical process and visualize the time series decomposition. 

import plotly

# Plotly plots will be rendered as PNG images, uncomment the following line to crate an interactive chart
plotly.io.renderers.default = 'png'
eda.plot_model(plot="decomp")
classical_decomposition_time_series
 
 By using the plot_model function, PyCaret generates decomposition graphs, displaying the trend, seasonality, and residuals of our time series:
 
eda.plot_model(plot="decomp", data_kwargs={'seasonal_period': 20})
classical_decomposition_time_series2
We can adjust the seasonality period and employ more advanced decomposition methods, such as the STL decomposition, for enhanced flexibility.
 
eda.plot_model(plot="decomp_stl")
 
stl_decomposition_time_series
 Step 8. Analyzing Statistical Properties with PyCaret

PyCaret's check_stats function provides a comprehensive statistical snapshot of our time series data, performing a suite of statistical tests. These tests provide insights into the central tendency, dispersion, and shape of the dataset's distribution.
 
eda_check_stats
 

With this overview of time series decomposition with PyCaret and CrateDB, you now have a solid understanding of the data’s statistical properties which is essential for reliable and insightful anomaly detection and forecasting.

Want to read more?

Download now

Whitepaper" Guide for Time Series Data Projects

Related resources