Seasonality

What is Seasonality in Time Series Data?

A time series data is said to be seasonal if specific patterns appear on a cyclical basis. The cycle repeating over time may signal important information in forecasting.

A cycle structure in a time series may or may not be seasonal. If it consistently repeats at the same frequency, 
it is seasonal, otherwise it is not seasonal and is called a cycle.
- Page 6, Introductory Time Series with R

Types of Seasonality

There can be many types of seasonality based on their repeating cycle. For example:

• Hourly
• Daily
• Weekly
• Monthly
• Quarterly
• Annually

Why Do We Need to Detect Seasonality?

1. Understanding the seasonal component can help in improving the performance of the machine learning model.

How to Detect Seasonality?

The most popular method to detect seasonality components is decomposing the time series data.

Time Series Decomposition

The time series data is decomposed into three components: trend, seasonal and residual.

1. Trend

The consistently increasing or decreasing values in a series. It is used to find trend in the time series data under consideration i.e uptrend or downtrend.

2. Seasonal

The repeating short-term cycle in the series. It is used to find seasonality in the data.

3. Residual

If the trend or seasonality component is removed from a time series data, whatever is left over is called residual.

Naive Decomposition

In Naive Decomposition, we find the linear relationship between the three components (trend, seasonal, residual) and the observed time series. There are two main approaches to naive decomposition: additive and multiplicative.

1. Additive Decomposition

from random import randrange

from matplotlib import pyplot

from statsmodels.tsa.seasonal import seasonal_decompose

series=[i+randrange(10) for i in range(1,100)]

result=seasonal_decompose(series,model="additive",period=1)

result.plot()

pyplot.show()

observed = trend + seasonality + residual

2. Multiplicative Decomposition

series = [i*2 for i in range(1,100)]

result= seasonal_decompose(series, model="multiplicative",period=1)
result.plot()

pyplot.show()

observed = trend * seasonality * residual

How to Remove Seasonality?

The process of removing the seasonality from a time series data is called seasonal adjustment or deseasonalizing.

A time series data with removed seasonality is called seasonal stationary data.

One of the popular methods for seasonal adjustment is differencing.

Differencing

In this method, the level of data (weekly, monthly, yearly ) is found. The current value is subtracted from the level of data. For example, if the weekly data is available, then every value is subtracted by 7.