Forecasting with Artificial Intelligence : Theory and Applications

• Autore: Hamoudia Mohsen
• Pubblicazione: Cham : , : Palgrave Macmillan, , 2023; ©2023
• Edizione: 1st ed.
• Descrizione fisica: 1 online resource (441 pages)
• Disciplina: 003.2
• Soggetto topico: Artificial intelligence; Forecasting
• Altri autori: MakridakisSpyros SpiliotisEvangelos
• Nota di contenuto: Intro -- Foreword -- Preface -- Acknowledgments -- Contents -- Editors and Contributors -- About the Editors -- Contributors -- Abbreviations -- List of Figures -- List of Tables -- Part I Artificial Intelligence: Present and Future -- 1 Human Intelligence (HI) Versus Artificial Intelligence (AI) and Intelligence Augmentation (IA) -- Introduction -- Defining Intelligence -- The Distinctive Nature of AI and HI -- The Evolution of HI and Its Complementarity with AI -- AI: Current Capabilities and Future Challenges -- The Future of AI: Uncertainties and Challenges -- AVs: Achievements, Limitations, and Future Prospects -- The Current Situation -- The Four AI/HI Scenarios -- Smarter, Faster Hedgehogs -- An Army of Specialized Hedgehogs -- Fully Integrated HI and AI -- More Efficient Ways of Communicating with Computers -- Direct Brain to Computer Interface (BCI) -- Non-Invasive Forms of Interfaces -- Invasive Forms of Interfaces -- Brain-To-Brain Interfaces (BBI) -- The Future World of Transhumanism -- Runaway AGI -- Conclusions -- References -- 2 Expecting the Future: How AI's Potential Performance Will Shape Current Behavior -- Introduction -- The Black Box of Machine Learning -- Decreased Efforts -- Educational Choices -- Political Job Protection -- Greater Desire of Governments to Borrow -- Economies of Scale -- Trade Blocks -- What Happens If People Expect Money to Become Worthless -- Animal Welfare -- Conclusion -- References -- Part II The Status of Machine Learning Methods for Time Series and New Product Forecasting -- 3 Time Series Forecasting with Statistical, Machine Learning, and Deep Learning Methods: Past, Present, and Future -- Introduction -- Statistical Methods -- Machine Learning Methods -- Deep Learning Methods -- Discussion and Future Steps -- References -- 4 Machine Learning for New Product Forecasting -- Introduction. ; Categories of New Products and Data Availability -- Data Availability -- There is Data on Similar Products that Have Been Adopted in the Past -- The Product is Truly New -- Highlights of the Main Forecasting Techniques Used for New Products -- "Traditional" Methods -- Expert Opinion -- The Delphi Method -- Consensus of Sales Team -- Customers Surveys -- Drivers and Constraints -- Diffusion Models (S-Shaped Curves) -- Machine Learning Methods for Predicting New Product Demand -- Gradient Boosted Trees (GBT) -- Artificial Neural Network (ANN) -- Structuring the New Product Forecasting Problem -- Potential Machine Learning Approaches for Truly New-to-the-World Products -- A Review of Four Case Studies of Machine Learning for New Product Forecasting -- Forecasting Demand Profiles of New Products -- Description -- Test Setup -- Results -- Feature Importance and Comparable Products -- Summary and Conclusion -- Fashion Retail: Forecasting Demand for New Items -- Methodology -- Experiment and Results -- Conclusion and Future Directions -- New Product Forecasting Using Deep Learning - A Unique Way -- Methodology and Results -- Summary and Conclusion -- Forecasting New Product Demand Using Machine Learning -- Methods -- Results -- Summary and Conclusion -- Summary and Lessons from the Four Case Studies -- Summary and Conclusions -- References -- Part III Global Forecasting Models -- 5 Forecasting with Big Data Using Global Forecasting Models -- Big Data in Time Series Forecasting -- Global Forecasting Models -- A Brief History of Global Models -- Model Complexity in Global Models -- Global Forecasting Model Frameworks -- Recent Advances in Global Models -- Conclusions and Future Work -- References -- 6 How to Leverage Data for Time Series Forecasting with Artificial Intelligence Models: Illustrations and Guidelines for Cross-Learning -- Introduction. ; Background and Definitions -- Common Terminology and Building Blocks -- Locality and Globality -- Approximation Capacity of Cross-Learned Models -- Statistical Trade-Offs -- Illustration of Cross-Learning Approximation of Specific Time Series Processes -- Well-Known Families of Time Series Processes Have an Equivalent Parameterization via Local Autoregression -- Two Different Linear Autoregressive Predictive Functions Can Be Represented by a Single Linear Autoregressive with More Lags -- A Single Nonlinear Autoregression Can Represent an Arbitrary Number of Different Linear Autoregressions -- Cross-Learning for Processes with Noise or Perturbations -- Explicit Example of Transfer Learning -- Conclusions -- References -- 7 Handling Concept Drift in Global Time Series Forecasting -- Introduction -- Problem Statement: Concept Drift Types -- Related Work -- Global Time Series Forecasting -- Transfer Learning -- Concept Drift Handling -- Methodology -- Series Weighted Methods -- Our Proposed Concept Drift Handling Methods -- Error Contribution Weighting -- Gradient Descent Weighting -- Experimental Framework -- Datasets -- Error Metrics -- Evaluation -- Benchmarks -- Statistical Testing of the Results -- Results and Discussion -- Main Accuracy Results -- Statistical Testing Results -- Further Insights -- Conclusions and Future Research -- References -- 8 Neural Network Ensembles for Univariate Time Series Forecasting -- Introduction -- Experimental Design -- Initialization Seed -- Loss Function -- Input Layer Size -- Empirical Evaluation -- Time Series Data -- Forecasting Performance Measures -- Base Forecasting Neural Network -- Results and Discussion -- Initialization Seed -- Loss Function -- Input Layer Size -- Final Ensemble -- Conclusions -- References -- Part IV Meta-Learning and Feature-Based Forecasting. ; 9 Large-Scale Time Series Forecasting with Meta-Learning -- Introduction -- A Brief Review of the Meta-Learning in Time Series Forecasting -- Key Choices in Designing an Effective Meta-Learning Framework -- Selection of Base Forecasters -- Feature Extracting -- Meta-Learners -- A Python Library for Time Series Forecasting with Meta-Learning -- Raw Data Class -- Metadata Class -- Meta-Learner -- The Meta-Combination Learner with Encapsulated Features Extractor ('MetaComb') -- The Meta-Combination Learner with Unsupervised Hand-Selected Features ('MetaCombFeat') -- The Model Selection Learner Targets Identifying the Best Base Forecaster ('MetaSelection') -- The Metaloss Prediction Learner ('MetaLoss') -- Experimental Evaluation -- Datasets -- Predictive Performance -- Conclusions and Future Directions -- References -- 10 Forecasting Large Collections of Time Series: Feature-Based Methods -- Introduction -- Data Generation -- Diverse Time Series Generation -- Time Series Generation with Target Features -- Feature Extraction -- Time Series Features -- Automation of Feature Extraction -- Time Series Imaging -- Forecast Diversity -- Automatic Feature Selection -- Forecast Trimming for Combination -- Accuracy, Robustness, and Diversity -- Forecast Trimming -- Some Practical Forecasting Issues -- Intermittent Demand -- Uncertainty Estimation -- Conclusion -- References -- Part V Special Applications -- 11 Deep Learning Based Forecasting: A Case Study from the Online Fashion Industry -- Introduction -- Data for Forecasting at Zalando: An Overview -- Sales to Demand Translation -- Description of Covariates -- Demand Forecast Model -- Problem Formalization -- Model Architecture -- Input Preparation -- Encoder -- Positional Encoding -- Padding and Masking -- Decoder -- Monotonic Demand Layer -- Near and Far Future Forecasts -- Training -- Prediction. ; Empirical Results -- Accuracy Metrics -- Model Benchmarking -- On the Benefits of Transformer-based Forecasting: First Results -- Related Work -- Practical Considerations -- Conclusion -- References -- 12 The Intersection of Machine Learning with Forecasting and Optimisation: Theory and Applications -- Introduction -- Predict and Optimise -- Problem Setting -- Methodologies and Applications -- Discussion and Future Works -- Predicting the Solutions of Optimisation Problems -- Problem Setting -- Methodologies and Applications -- Discussion and Future Works -- Conclusion -- References -- 13 Enhanced Forecasting with LSTVAR-ANN Hybrid Model: Application in Monetary Policy and Inflation Forecasting -- Introduction -- LSTVAR-ANN Hybrid Model -- LSTVAR Model -- LSTAR-ANN Hybrid Model -- Research Architecture Specification -- 3D Impulse Response Function -- Monetary Policy Model with Regimes -- Data -- Results -- Unobserved Factors Extraction -- Importance on CPI Prediction -- Transition Function Output -- Response of CPI on Interest Rate Shock -- Forecasting -- Pre-COVID-19 Period -- COVID-19 and Post-COVID-19 Period -- Conclusion -- Appendix -- References -- 14 The FVA Framework for Evaluating Forecasting Performance -- What is FVA? -- Why Use FVA? -- Applying the Scientific Method to Forecasting -- FVA Analysis: Step-by-Step -- Mapping the Process -- Collecting the Data -- Analyzing the Process -- Reporting the Results -- Interpreting the Results -- FVA Challenges and Extensions -- Compete or Combine? -- Stochastic Value Added (SVA): FVA for Probabilistic Forecasts -- Summary -- References -- Author Index -- Subject Index.
• Sommario/riassunto: This book explores the intersection of Artificial Intelligence (AI) and forecasting, providing an in-depth analysis of AI's current capabilities and its implications for forecasting theory and practice. Comprising 14 chapters, it covers topics such as the impact of AI on traditional and machine forecasting methods, challenges in demand forecasting, meta-learning, and feature-based forecasting. The book discusses AI's role in improving forecasting accuracy and addresses scalability and computational efficiency issues. It also examines the use of ensemble learning techniques to enhance performance. Designed for researchers and professionals in the field of time series forecasting, the book combines theoretical insights with practical applications, offering valuable perspectives on leveraging AI to improve forecasting processes.
• Titolo autorizzato: Forecasting with Artificial Intelligence
• ISBN: 9783031358791; 3031358791
• Formato: Materiale a stampa
• Livello bibliografico: Monografia
• Lingua di pubblicazione: Inglese
• Record Nr.: 9910746292703321
• Lo trovi qui: Univ. Federico II
• Serie: Palgrave Advances in the Economics of Innovation and Technology Series