SAS Institute Of Management Studies - Modeling Time Series and Sequential Data 

Overview

Getting the Most from this Specialization

Welcome to the course

Finding the Course Files and Practicing in this Course (REQUIRED)

Prerequisites

About this Module

Time Series Components

Applications of Time Series Analysis

Demo: Exploring a Time Series

A Framework for Forecasting

Demo: Accumulating a Time Series and Exploring Systematic Variation

Concepts and Notation

Naive Models

Introduction to Exponential Smoothing Models (ESM)

ESM and Signal Components

Demo: Forecasting with ESM

Question: Statistical Time Series

Practice: Forecasting with ESMs

About this Module

Models for Stationary Data

Autoregressive Moving Average Models

Identifying ARMA Models (Part 1)

Identifying ARMA Models (Part 2)

Demo: ARMA Model Properties

Automatic Order Identification

Demo: Identifying ARMA Orders

Non-Stationary Data, Trend

Differencing and Integration

Trend Functions

Demo: Trend Two Ways in an ARIMA Framework

The Augmented Dickey Fuller Unit Root (ADF) Test (Part 1)

The Augmented Dickey Fuller Unit Root (ADF) Test (Part 2)

Demo: An Application of the ADF Test

Seasonal Variation (Part 1)

Seasonal Variation (Part 2)

The ADF Test for Seasonality

Demo: Seasonality Two Ways in an ARIMA Framework

Time Series Regression

Demo: Ordinary Regression Using Outliers

The Cross Correlation Function (CCF)

The Transfer Function

Interpreting the CCF

Demo: Dynamic Regression with Event Variables

Cross Correlation Pitfalls

Question: Stationary Time Series

Practice: ARIMAX - Identification, Estimation and Forecasting

About this Module

Classical Analysis versus Bayesian Analysis

Accessing Lag and Next Values

Demo: Setting Up Autoregressive Components

Dynamic Linear Model Setup

Demo: Setting Up Seasonality Components

Adding Exogenous Variables

Demo: Setting Up Exogenous Components

PREDDIST and Forecasting

Demo: Forecast Output

Question: PROC MCMC Diagnostics

Question: PROC MCMC Statements

Practice: Modeling Autoregressive Components in Concert Data

Practice: Modeling Seasonality Components in Stock Data

Question: PROC MCMC Syntax

Practice: Modeling Exogenous Components in Rose Sales Data

Question: Bayesian Scoring/Forecasting Techniques

Practice: Generating Posterior Predictive Distributions for an AR(1) Model

About this Module

Preparing Time Series Data for Machine Learning

Brief Introduction to Gradient Boosting Models

Demo: Preparing Time Series Data and Building a Gradient Boosting Model"

Introduction to Recurrent Neural Networks

Long Short-Term Memory Blocks in RNNs

Demo: Building a Recurrent Neural Network with LSTM Blocks to Forecast Time Series

Limitations of Machine Learning Methods for Time Series Forecasting

About the Next Three Practices

Question: Machine Learning Models

Question: Data for Recurrent Neural Network Modeling

Practice: Changing the Number of Lagged Input Values for the Recurrent Neural Network Model

Practice: Adding Hidden Units to the Recurrent Neural Network Model

Practice: Removing a Hidden Layer from the Recurrent Neural Network Model

About this Module

External Models and Combined Forecasts

Combination Forecast Details

Combined Forecasts Using the TSM Package

Demo: Generating Combined Forecasts with the CFC Object

Demo: Combining Forecasts from Multiple Modeling Approaches

Strengths of Machine Learning Methods: Modeling Multiple Time Series

Weighting Combined Forecasts with Machine Learning

Demo: Using Gradient Boosting to Find the Best Weighted Combination of Traditional Time Series Models

Practice: Generating a Combined Model Forecast

Modeling Time Series and Sequential Data - Course Exam