Applied Time Series Analysis in Python

Name: Applied Time Series Analysis in Python
Rating: 4.3 (823 reviews)

Use Python and Tensorflow to apply the latest statistical and deep learning techniques for time series analysis

Created byMarco Peixeiro

Last updated 7/2022

English

What you'll learn

Descriptive vs inferential statistics
Random walk model
Moving average model
Autoregression
ACF and PACF
Stationarity
ARIMA, SARIMA, SARIMAX
VAR, VARMA, VARMAX
Apply deep learning for time series analysis with Tensorflow
Linear models, DNN, LSTM, CNN, ResNet
Automate time series analysis with Prophet

Course content

8 sections • 43 lectures • 6h 56m total length

Introduction2:58
Course introduction and overview
What are Time Series?1:46
Understand what time series are, understand their components, and gain an intuition of how to approach time series analysis.

Basic Statistics4:18
Learn about descriptive and inferential statistics
Setup for coding exercises0:11
Coding Exercise: Descriptive and Inferential Statistics20:50
Cover the basic statistics concepts in Python that will be needed throughout the course
Autocorrelation and White Noise2:17
Learn about autocorrelation, white noise, and the autocorrelation function (ACF)
Stationarity and Differencing1:45
Learn the concepts of sationarity and differencing, and see how it impacts the ACF plot.

Random Walk1:38
Learn about the random walk model and its ACF.
Coding Excercise: Random Walk8:02
Apply the random walk in Python
Moving Average Model1:23
Learn about the moving average model, and how to recognize its order from the ACF plot.
Coding Exercise: Moving Average Model8:29
Simulate a moving average process of order 2 in Python, and model the simulation to see how well we can approximate the weights at each lag.
Autoregressive Model2:24
Learn about the autoregressive model and about the partial autocorrelation function (PACF). You will see how the ACF and PACF plots look like when an autoregressive process is in play.
Mini Project: Autoregressive Model20:30
Simulate AR processes and apply the AR model in a mini project to model the quarterly earnings per share of Johnson&Johnson
ARMA: Autoregressive Moving Average Model'1:37
Combine the MA(q) and AR(p) models into an ARMA(p,q) model and learn to recognize this process by analyzing the ACF and PACF.
Coding Exercise: ARMA7:00
Simulate an ARMA process and evaluate the ACF and PACF plots

ARIMA: Autoregressive Integrated Moving Average Model1:28
Learn the theory of ARIMA and gain an intuition of how ARIMA works
Project 1: ARIMA24:40
Revisit the Johnson&Johnson quarterly EPS dataset and apply an ARIMA model. Learn how to optimize it, how to make predictions, plot forecast, and evaluate the model.
SARIMA1:53
Add another layer of complexity to ARIMA and learn about SARIMA, which now considers the seasonality of a dataset.
Project 2: SARIMA23:45
Optimize and apply the SARIMA model to the Johnson&Johnson Quarterly EPS dataset.
AIC: Akaike Information Criterion1:38
Learn about the AIC (Akaike's information criterion) which will help us select the most suitable model for our projects.
SARIMAX1:04
Add a final layer of complexity and learn about SARIMAX, which allows us to consider multiple variables when forecasting our time series.
Project 3: SARIMAX30:26
Apply the SARIMAX model to forecast the US GDP.
General Modelling Procedure0:58
Here, we outline a general modelling procedure that can be applied to any projects involving the analysis of a time series.
VAR: Vector Autoregressions1:29
Learn how to forecast multivariate time series using VAR.
Project 4 - Part 1: VAR23:37
Apply the VAR model to predict a vector of 2 time series
Project 4 - Part 2: VARMA10:34
Apply the VARMA model to forecast a vector of 2 time series
Project 4 - Part 3: VARMAX14:10
Apply the VARMAX model to forecast a vector of 2 time series

Introduction1:10
Overview of what will be covered in the deep learning section.
Deep Neural Networks (DNN)0:48
Quick introduction on deep neural networks and how they can be used for time series analysis
Recurrent Neural Network and Long Short-Term Memory (RNN and LSTM)4:26
Learn the inner workings of RNNs and LSTMs
Convolutional Neural Networks (CNN)3:26
Learn the inner workings of a CNN architecture and how it can be useful for time series analysis

Project 5 - Part 1: Initial setup9:37
Import the required libraries and read in the dataset
Project 5 - Part 2: Exploratory Data Analysis (EDA)2:44
Analyze the features and remove those that will negatively impact our models
Project 5 - Part 3: Feature Engineering13:34
Perform feature engineering on the time variable
Project 5 - Part 4: Data Windowing and Training Function24:13
Implement a data windowing class and the training function that will be used throughout the entire project
Project 5 - Part 5: Single Step Models28:26
Run a series of experiments with single step models; predict 1 step into the future.
Project 5 - Part 6: Multi Output Models18:19
Run experiments on multi output models; predict multiple targets 1 step in the future
Project 5 - Part 7: Multi Step Models35:21
Conclude the project by running experiments on multi step models: predict many steps into the future.

Introduction to Prophet2:13
Learn about what Prophet is, what it can do, and follow the installation steps to get it ready on your machine!
Working with Prophet30:34
Download the dataset and start using Prophet! We explore the different functionalities of Prophet and apply them step by step.
Project: Predict Bus Ridership with Prophet18:36
Consolidate your learning of Prophet with this project, where we will predict bus ridership in Portland, Oregon! We get to apply Prophet on monthly data and include holiday effects, which we haven't covered yet!

Requirements

Basic knowledge of Python
Basic knowledge of deep learning
Jupyter notebook installed (or access to Google Colab)

Description

This is the only course that combines the latest statistical and deep learning techniques for time series analysis. First, the course covers the basic concepts of time series:

stationarity and augmented Dicker-Fuller test
seasonality
white noise
random walk
autoregression
moving average
ACF and PACF,
Model selection with AIC (Akaike's Information Criterion)

Then, we move on and apply more complex statistical models for time series forecasting:

ARIMA (Autoregressive Integrated Moving Average model)
SARIMA (Seasonal Autoregressive Integrated Moving Average model)
SARIMAX (Seasonal Autoregressive Integrated Moving Average model with exogenous variables)

We also cover multiple time series forecasting with:

VAR (Vector Autoregression)
VARMA (Vector Autoregressive Moving Average model)
VARMAX (Vector Autoregressive Moving Average model with exogenous variable)

Then, we move on to the deep learning section, where we will use Tensorflow to apply different deep learning techniques for times series analysis:

Simple linear model (1 layer neural network)
DNN (Deep Neural Network)
CNN (Convolutional Neural Network)
LSTM (Long Short-Term Memory)
CNN + LSTM models
ResNet (Residual Networks)
Autoregressive LSTM

Throughout the course, you will complete more than 5 end-to-end projects in Python, with all source code available to you.

Who this course is for:

Beginner data scientists looking to gain experience with time series
Deep learning beginners curious about times series
Professional data scientists who need to analyze time series
Data scientists looking to transition from R to Python

Applied Time Series Analysis in Python

What you'll learn

Explore related topics

Course content

Introduction2 lectures • 5min

Statistical Learning Approach: The Building Blocks5 lectures • 29min

Statistical Learning Approach: Basic Models8 lectures • 51min

Statistical Learning Approach: Advanced Models12 lectures • 2hr 16min

Deep Learning Approach: Theory4 lectures • 10min

Deep Learning Approach: End-to-end Project7 lectures • 2hr 12min

Conclusion and References2 lectures • 2min

Bonus: Automated Time Series Analysis with Prophet3 lectures • 51min

Requirements

Description

Who this course is for: