Introduction to Time Series Analysis and ARIMA

What is Time Series Analysis?

A time series is a set of observations collected or recorded over time at regular intervals.

Examples:

• Daily temperature readings
• Monthly sales of a company
• Annual population growth
• Daily stock prices
• Monthly rainfall
• Electricity consumption over time

For example:

Month	Sales
January	120
February	135
March	128
April	150

Since the observations are arranged according to time, this data is called a time series.

Time Series Analysis

Time Series Analysis is the process of analyzing data collected over time to understand its patterns and to make predictions about future values.

The main objectives are:

• To identify patterns and trends.
• To understand seasonal variations.
• To study changes over time.
• To forecast future values.

---

Components of a Time Series

A time series may contain the following components:

1. Trend

A trend represents the long-term direction of the data.

For example:

The sales of a company gradually increase every year.

2. Seasonality

Seasonality refers to patterns that repeat at regular intervals.

For example:

Ice cream sales increase every summer.

3. Cyclical Variation

These are long-term fluctuations that may occur over several years.

For example:

Changes in economic activity during business cycles.

4. Irregular Variation

These are random and unpredictable changes.

For example:

• Natural disasters
• अचानक changes in demand
• Unexpected events

---

What is ARIMA?

ARIMA is one of the most widely used models for time series forecasting.

ARIMA stands for:

A – AutoRegressive
I – Integrated
MA – Moving Average

An ARIMA model is generally represented as:

$$\boxed{ARIMA(p,d,q)}$$

where:

• p → Order of the AutoRegressive part
• d → Degree of differencing
• q → Order of the Moving Average part

---

1. AR – AutoRegressive

The AutoRegressive (AR) component uses previous values of the time series to predict the current value.

For example:

Today's sales may depend on the sales of the previous few days.

The parameter p indicates how many previous observations are used.

---

2. I – Integrated

The Integrated (I) component refers to differencing the data to make the time series stationary.

The parameter d represents the number of times differencing is performed.

For example:

Original values

100, 110, 125, 140

First Difference

10, 15, 15

Differencing helps remove trends and makes the data more suitable for ARIMA modelling.

---

3. MA – Moving Average

The Moving Average (MA) component uses previous forecast errors to predict the current value.

The parameter q indicates the number of previous error terms considered.

---

Understanding ARIMA(p, d, q)

For example:

$$ARIMA(2,1,1)$$

means:

Parameter	Value	Meaning
p	2	Uses 2 previous observations
d	1	Data is differenced once
q	1	Uses 1 previous forecast error

---

Why is Stationarity Important?

Most traditional time series models, including ARIMA, work best when the time series is stationary.

A stationary time series generally has:

• Constant mean
• Constant variance
• Stable statistical behaviour over time

If the data contains a strong trend, we often use differencing to make it stationary.

---

ARIMA in R

R provides built-in functions for time series analysis.

A time series can be created using:

ts()

An ARIMA model can be created using:

arima()

For example:

model <- arima(data, order = c(p, d, q))

---

Simple Example

model <- arima(sales, order = c(1, 1, 1))

This creates an:

$$ARIMA(1,1,1)$$

model.

The model can then be used to forecast future values.

---

Summary

Time Series Analysis deals with data collected over time and is used to identify patterns such as trend and seasonality and to predict future values. ARIMA is a popular time series forecasting model that combines AutoRegression, differencing (Integration), and Moving Average techniques. An ARIMA model is represented as ARIMA(p,d,q), where p represents the AutoRegressive order, d represents the degree of differencing, and q represents the Moving Average order.

Quick Comparison

Concept	Main Purpose
Time Series Analysis	Analyze data over time
Trend	Identify long-term movement
Seasonality	Identify repeating patterns
AR	Uses previous observations
I	Makes data stationary through differencing
MA	Uses previous forecast errors
ARIMA	Forecasts future values