book

Advanced Forecasting with Python: Mastering Modern Forecasting Techniques with Machine Learning and Cloud Tools

Name: Advanced Forecasting with Python: Mastering Modern Forecasting Techniques with Machine Learning and Cloud Tools
Author: Joos Korstanje
ISBN: 9798868820281

by Joos Korstanje

November 2025

Intermediate to advanced

435 pages

8h 12m

English

Apress

Read now

Unlock full access

Advanced Forecasting with Python
Introduction
Table of Contents
About the Author
About the Technical Reviewer
Part I: Machine Learning for Forecasting
1. Models for Forecasting
Reading Guide for This BookMachine Learning LandscapeUnivariate Time Series ModelsA Quick Example of the Time Series ApproachSupervised Machine Learning ModelsA Quick Example of the Supervised Machine Learning ApproachCorrelation CoefficientOther Distinctions in Machine Learning ModelsSupervised vs. Unsupervised ModelsClassification vs. Regression ModelsUnivariate vs. Multivariate ModelsKey Takeaways
2. Model Evaluation for Forecasting
Evaluation with an Example ForecastError MetricsError Metric 1: MSEError Metric 2: RMSEError Metric 3: R2Error Metric 4: MAEError Metric 5: MAPEModel Evaluation StrategiesOverfit and the Out of Sample ErrorStrategy 1: Train-Test SplitStrategy 2: Train-Validation-Test SplitStrategy 3: Cross-Validation for ForecastingK-Fold Cross-ValidationTime Series Cross-ValidationRolling Time Series Cross-ValidationBacktestingWhich Strategy to Use for Safe Forecasts?Final Considerations on Model EvaluationKey Takeaways
3. Model Management and Benchmarking Using MLflow
Introduction to MLflowLocal vs. Hosted MLflowSetting Up MLflow LocallyExperiment Tracking Using MLflowMLflow DataArtifactsMetricsParamsTagsInspecting the Model Logs Through PythonThe MLflow UIMLflow Throughout This BookKey Takeaways
Part II: Univariate Time Series Models

4. The AR Model
Autocorrelation: The Past Influences the PresentCompute Autocorrelation in Earthquake CountsPositive and Negative AutocorrelationStationarity and the ADF TestDifferencing a Time SeriesLags in AutocorrelationPartial AutocorrelationHow Many Lags to Include?AR Model DefinitionEstimating the AR Using Yule–Walker EquationsThe Yule–Walker MethodTrain, Test, Evaluation, and TuningSaving the Final Model Using MLflowKey Takeaways
5. The MA Model
The Model DefinitionFitting the MA ModelStationarityChoosing Between an AR and an MA ModelApplication of the MA ModelMultistep Forecasting with Model RetrainingGrid Search to Find the Best MA OrderSaving This Model in MLflowKey Takeaways
6. The ARMA Model
The Idea Behind the ARMA ModelThe Mathematical Definition of the ARMA ModelAn Example: Predicting Births Using ARMAFitting an ARMA(1,1) ModelMore Model Evaluation KPIsAutomated Hyperparameter TuningGrid Search: Tuning for Predictive PerformanceSaving This Model in MLflowKey Takeaways
7. The ARIMA Model
ARIMA Model DefinitionModel DefinitionARIMA on the CO2 ExampleKey Takeaways
8. The SARIMA Model
Univariate Time Series Model BreakdownThe SARIMA Model DefinitionExample: SARIMA on Walmart SalesKey Takeaways
Part III: Multivariate Time Series Models
9. The SARIMAX Model
Time Series Building BlocksModel DefinitionSupervised Models vs. SARIMAXExample of SARIMAX on the Walmart DatasetKey Takeaways
10. The VAR Model
The Model DefinitionOrder: Only One HyperparameterStationarityEstimation of the VAR CoefficientsOne Multivariate Model vs. Multiple Univariate ModelsAn Example: VAR for Forecasting Walmart SalesKey Takeaways
11. The VARMAX Model
Model DefinitionMultiple Time Series with Exogenous VariablesKey Takeaways
Part IV: Supervised Models
12. The Linear Regression
Linear RegressionModel DefinitionExample: Linear Model to Forecast CO2 LevelsKey Takeaways
13. The Decision Tree Model
MathematicsSplittingPruning and Reducing ComplexityExampleKey Takeaways
14. The kNN Model
Intuitive ExplanationMathematical Definition of Nearest NeighborsCombining k Neighbors into One ForecastDeciding on the Number of Neighbors kPredicting Traffic Using kNNGrid Search on kNNRandom Search: An Alternative to Grid SearchKey Takeaways
15. The Random Forest
Intuitive Idea Behind Random ForestsRandom Forest Concept 1: Ensemble LearningBagging Concept 1: BootstrapBagging Concept 2: AggregationRandom Forest Concept 2: Variable SubsetsPredicting Births Using a Random ForestGrid Search on the Two Main Hyperparameters of the Random ForestRandom Search CV Using DistributionsDistribution for max_featuresDistribution for n_estimatorsFitting the RandomizedSearchCVInterpretation of Random Forests: Feature ImportanceKey Takeaways
16. Gradient Boosting with XGBoost, LightGBM, and CatBoost
Boosting: A Different Way of Ensemble LearningGradient BoostingThe Difference Between XGBoost and LightGBMAdding CatBoost to the MixForecasting Traffic Volume with XGBoostForecasting Traffic Volume with LightGBMForecasting Traffic Volume with CatBoostInspecting the Differences in PredictionsHyperparameter Tuning Using Bayesian OptimizationThe Theory of Bayesian OptimizationBayesian Optimization Using scikit-optimizeConclusionKey Takeaways
17. Bayesian Models with pyBATS
Intuitive Idea Behind Bayesian ModelsBayesian vs. Frequentist StatisticspyBATS As an Implementation of Bayesian ForecastingForecasting Sales Using pyBATSSales Forecast: Exploratory Data AnalysisSales Forecast: Univariate Poisson DGLMSales Forecast: Normal DGLM with X VariableKey Takeaways
Part V: Neural Networks
18. Neural Networks
Fully Connected Neural NetworksActivation FunctionsThe Weights: BackpropagationOptimizersLearning Rate of the OptimizerHyperparameters at Play in Developing a NNIntroducing the Example DataSpecific Data Prep Needs for NNScaling and StandardizationPrincipal Component Analysis (PCA)The Neural Network Using KerasConclusionKey Takeaways
19. RNNs Using SimpleRNN and GRU
What Are RNNs: ArchitectureInside the SimpleRNN UnitThe ExamplePredicting a Sequence Rather Than a ValueUnivariate Model Rather Than MultivariablePreparing the DataA Simple SimpleRNNSimpleRNN with Hidden LayersSimple GRUGRU with Hidden LayersKey Takeaways
20. LSTM RNNs
What Is LSTMThe LSTM CellExampleLSTM with 1 Layer of 8LSTM with 3 Layers of 64ConclusionKey Takeaways
Part VI: Black Box and Cloud-Based Models
21. The N-BEATS Model with Darts
Intuition and Mathematics of N-BEATSInterpretabilityStackingAutomatic Feature Engineering and DecompositionThe Darts Package and Its Implementation of N-BEATSForecasting Sales Using N-BEATS in DartsSales Forecast: Preparing the DataSales Forecast: Create Default N-BEATS ModelSales Forecast: Create Multivariate N-BEATS ModelSales Forecast: Tuning the Multivariate N-BEATS ModelBenchmark ResultsKey Takeaways
22. The Transformer Model with Darts
Intuition and Mathematics of TransformersAttention Is All You NeedTransformers Move Away from Recurrent UnitsPositional EncodingSelf-AttentionThe Darts Package and Its Implementation of TransformersForecasting Sales Using Transformer in DartsSales Forecast: Preparing the DataSales Forecast: Create a Default N-BEATS ModelSales Forecast: Create the Multivariate Transformer ModelSales Forecast: Tuning the Multivariate Transformer ModelBenchmark ResultsKey Takeaways
23. The NeuralProphet Model
The NeuralProphet ModelPredicting Heat Waves with NeuralProphetWikipedia Pageview ToolPreparing the Data in PythonTime Series Decomposition Using NeuralProphetAdvanced Decomposition SettingTrain-Test SplitDefault ModelTuned ModelMLflowKey Takeaways
24. The DeepAR Model and AWS SageMaker AI
About DeepARForecasting Heat Wave Page Views Using gluonts DeepAR LocallyTuning the DeepAR Model Using Ray HyperOptSearchWhat Is Ray HyperOptSearch?DeepAR vs. AWS SageMaker AIForecasting Heat Wave Page Views Using DeepAR Inside an AWS SageMaker AI Training JobStep 1: Prepare the Data FileStep 2: Upload the Data to S3Step 3: SageMaker AIKey Takeaways
25. Uber’s Orbit Model
About OrbitForecasting Heat Wave Page Views Using Orbit’s Local-Global Trend ModelTuning Orbit’s GLTForecasting Heat Wave Page Views Using Orbit’s Damped Local Trend ModelTuning Orbit’s DLTKey Takeaways
26. AutoML with Microsoft Azure
Cloud ComputingFrom Cloud-Based Maths to AutoMLMicrosoft Azure CloudForecasting Heat Wave Page Views with Microsoft Azure Cloud AutoMLStep 1: Create an Azure Machine Learning StudioStep 2: Create an Automated ML RunOptional: Download MLflow Artifact of the ModelKey Takeaways
27. AutoML with Vertex AI on Google Cloud Platform
GCP and BigQueryVertex AIForecasting Heat Wave Page Views with Google Cloud Platform’s Vertex AI AutoMLStep 1: Data Preparation for GCPStep 2: Create a BucketStep 3: Uploading the DataOptional Next StepsKey Takeaways
28. Nixtla Suite and TimeGPT
The NixtlaVerseSimple Use Case with the NixtlaVerseThe GPT in TimeGPTNixtla’s TimeGPT APIKey Takeaways
29. Model Selection
Model Selection Based on MetricsModel Structure and InputsOne-Step Forecasts vs. Multistep ForecastsModel Complexity vs. GainModel Complexity vs. InterpretabilityModel Stability and VariationOpen Source vs. Cloud Services Black BoxConclusionKey Takeaways
Index

Content preview from Advanced Forecasting with Python: Mastering Modern Forecasting Techniques with Machine Learning and Cloud Tools

J. KorstanjeAdvanced Forecasting with Pythonhttps://doi.org/10.1007/979-8-8688-2028-1_16

16. Gradient Boosting with XGBoost, LightGBM, and CatBoost

Joos Korstanje¹

(1)

Maisons Alfort, France

In this chapter, you will discover the Gradient Boosting model. In the previous chapter, you discovered the idea behind ensemble methods. As a recap, ensemble methods make powerful predictions by combining predictions of numerous small, less performant models.

Boosting: A Different Way of Ensemble Learning

Gradient Boosting combines numerous small decision tree models to make predictions. Of course, those small decision trees are different from each other: otherwise, there ...

Become an O’Reilly member and get unlimited access to this title plus top books and audiobooks from O’Reilly and nearly 200 top publishers, thousands of courses curated by job role, 150+ live events each month,
and much more.

Read now

Unlock full access

More than 5,000 organizations count on O’Reilly

O’Reilly covers everything we've got, with content to help us build a world-class technology community, upgrade the capabilities and competencies of our teams, and improve overall team performance as well as their engagement.

Julian F.

Head of Cybersecurity

I wanted to learn C and C++, but it didn't click for me until I picked up an O'Reilly book. When I went on the O’Reilly platform, I was astonished to find all the books there, plus live events and sandboxes so you could play around with the technology.

Addison B.

Field Engineer

I’ve been on the O’Reilly platform for more than eight years. I use a couple of learning platforms, but I'm on O'Reilly more than anybody else. When you're there, you start learning. I'm never disappointed.

Amir M.

Data Platform Tech Lead

I'm always learning. So when I got on to O'Reilly, I was like a kid in a candy store. There are playlists. There are answers. There's on-demand training. It's worth its weight in gold, in terms of what it allows me to do.

Mark W.

Embedded Software Engineer

Modern Time Series Forecasting with Python

Publisher Resources

ISBN: 9798868820281Purchase Link Publisher Website

Cloud Computing

Data Engineering

Data Science

AI & ML

Programming Languages

Software Architecture

IT/Ops

Security

Design

Business

Soft Skills

Advanced Forecasting with Python: Mastering Modern Forecasting Techniques with Machine Learning and Cloud Tools

by Joos Korstanje

16. Gradient Boosting with XGBoost, LightGBM, and CatBoost

Boosting: A Different Way of Ensemble Learning

Become an O’Reilly member and get unlimited access to this title plus top books and audiobooks from O’Reilly and nearly 200 top publishers, thousands of courses curated by job role, 150+ live events each month,
and much more.