Scribd
Upload
13 views4 pages

TSA Chapters 1: introduction to time series

This document introduces time series, defining it as a sequence of data points collected at successive time intervals, and outlines its key components such as trend, seasonality, and random fluctuations. It emphasizes the importance of time series forecasting in various domains and categorizes forecasting methods into quantitative and qualitative approaches, including advanced techniques like ARIMA and machine learning models. Additionally, it provides guidance on choosing the right forecasting method based on the nature of the data and accuracy requirements.

Uploaded by

oumaima abaied
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
13 views4 pages

TSA Chapters 1: introduction to time series

This document introduces time series, defining it as a sequence of data points collected at successive time intervals, and outlines its key components such as trend, seasonality, and random fluctuations. It emphasizes the importance of time series forecasting in various domains and categorizes forecasting methods into quantitative and qualitative approaches, including advanced techniques like ARIMA and machine learning models. Additionally, it provides guidance on choosing the right forecasting method based on the nature of the data and accuracy requirements.

Uploaded by

oumaima abaied
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 4

Chapter 1: introduction

Definition of Time Series


A time series is a sequence of data points collected or recorded at successive time intervals. The key
characteristic of a time series is its ordering, which typically corresponds to time but could also represent
other ordered dimensions (e.g., spatial sequences). In this course, we will focus on time series that:
●​ Are measured on a single continuous random variable (Y).
●​ Are equally spaced in discrete time, meaning we observe a single realization of Y at each fixed
interval (e.g., second, hour, day, month, year).

Components of a Time Series


Time series data typically consist of the following components:
●​ Trend: A long-term persistent increase or decrease in data values over time.
●​ Seasonality: Recurring patterns or fluctuations in data at regular intervals (e.g., sales peaking
every December).
●​ Cyclical Component: Long-term fluctuations that do not follow a fixed period, often influenced by
economic cycles.
●​ Random (Noise) Fluctuations: Irregular variations that do not follow any discernible pattern,
usually short-lived and unpredictable.

Time Series Forecasting


Time series forecasting is the process of predicting future values based on previously observed data.
The goal is to create models that capture patterns in past data and apply them to make informed
decisions.
Importance of Forecasting
Time series forecasting is crucial in multiple domains, including:

★​ Strategic Planning: Long-term decision-making and business strategy.


★​ Finance & Accounting: Budgeting, cost control, and financial planning.
★​ Marketing: Forecasting sales, market trends, and new product demand.
★​ Operations & Production: Inventory management, scheduling, and resource allocation.
Types of Forecasting Methods
Forecasting methods can be broadly classified into quantitative and qualitative approaches:
Quantitative Forecasting Methods (Data-Driven)
These methods rely on numerical data and analytical techniques. Common approaches include:
Causal Models:
Identify relationships between variables and time series patterns.
Time Series Models:
○​ Stationary Models: Assume constant statistical properties over time.
○​ Trend Models: Capture long-term upward or downward movement.
○​ Trend + Seasonality Models: Account for both trend and seasonal fluctuations.
Qualitative Forecasting Methods (Expert-Driven)
These methods rely on subjective opinions and expert judgment, often used when data is scarce:
●​ Expert Judgment: Decisions based on professional experience.
●​ Delphi Method: Consensus-based forecasting by gathering opinions from multiple experts.
●​ Grassroots Forecasting: Aggregating input from individuals directly involved in operations.
Advanced Time Series Forecasting Methods
ARIMA (AutoRegressive Integrated Moving Average):
A widely used statistical model for univariate time series.

Classical Exponential smoothing:


statistical Averages past observations with exponentially decreasing weights.
models Used for data that do not have any obvious trends or seasonality.

VAR (Vector AutoRegression):


Used for multivariate time series forecasting.

MACHINE
LEARNING

LSTM (Long
Short-Term
Memory
Networks):
Overcome the
Advanced RNN (Recurrent Neural vanishing gradient
Time Series Networks): problem in long
Capture sequential sequences.
Forecasting
dependencies in time
GRU (Gated
series.
Advanced ML Recurrent Units):

models: A simpler
alternative to
DEEP LEARNING
LSTMs with
comparable
performance.

CNN(Convolutional
Neural Networks):
Extracts spatial and
temporal features from
time series data.

TRANSFORMER:
Leverage self-attention
mechanisms for improved
sequential forecasting.
Choosing the Right Forecasting Method
When selecting a forecasting method, consider:

●​ The nature of the time series: Is it stationary, trending, or seasonal?


●​ The amount of available data: Some methods require large datasets.
●​ The complexity of the data: Nonlinear relationships may require advanced techniques like deep
learning.
●​ The accuracy requirements: More complex models tend to offer higher accuracy but require greater
computational resources.

By understanding these concepts, you can effectively analyze time series data and select the most
appropriate forecasting approach for your needs

Python for TSA:


Libraries:
-​ Pandas: preprocess data
-​ Matplots: visualize

You might also like