WEITS: A wavelet-enhanced residual framework for interpretable time series forecasting

Abstract: This paper assesses the contemporaneous relationship between stock prices, earnings and long‐term government bond yields for a large number of countries. The time period of our data spans several decades. In a time series framework our analysis first tests the presence of a long‐term contemporaneous relationship between these three variables (the so‐called Fed model). Next, we assess if government bond yields play a significant role in the long‐run relationship. Our empirical results question the validity of the Fed model in the sense that we show that long‐term market movements are mainly driven by the earnings yield and not the differential between bond and earnings yields. As such, our analysis validates the results of Asness (2003) for a much larger collection of countries while using a dynamic time series (cointegration) framework. Finally, we also show that changes in long‐term government bond yields have a short‐term impact on stock prices.

With an unparalleled number of Machine-to-Machine (M2M) devices being deployed to support a variety of smart-world systems powered by Internet of Things (IoT) technologies, the heterogeneity, uncertainty, and complexity of M2M communications have increased enormously. Thus, how to conduct network resource planning (NRP) has become a challenging issue. In this paper, we propose a novel time series framework (TSF) to model and forecast timescale network traffic dynamics in M2M communications that is capable of providing useful guidance for effective NRP. Specifically, our TSF utilizes the statistical techniques INGARCH(p,q) (integer valued generalized autoregressive conditional heteroskedasticity) and βARMA(p,q) (beta autoregressive moving average) to accurately capture both the internal and external impact factors of the asynchronous and synchronous M2M traffic dynamics over a large time scale, and produces forecasts for multiple upcoming time points by leveraging conditional maximum-likelihood estimators (CMLE). Through a combination of theoretical analysis and extensive simulation, we have validated the modeling and forecasting efficacy of TSF. Our experimental results demonstrate that TSF achieves superior performance with respect to goodness-of-fit and prediction accuracy.

Due to the time property of natural phenomena and human activities, time series are very common in our lives. The analysis and study of time series can help us to better understand the world, predict the future and make scientific decisions. Focusing on time series prediction, in this paper we propose a method of constructing non-periodic time series into periodic time series and design a framework for time series prediction based on the constructed periodic time series. The proposed construction method and prediction framework for the periodic time series are then applied to predict the state of health (SOH) of lithium-ion (Li-ion) batteries. The effectiveness of the proposed approach is verified and evaluated on publicly available datasets from the National Aeronautics and Space Administration (NASA), Ames Prognostics Center of Excellence (PCoE), and Center for Advanced Life Cycle Engineering (CALCE) of University of Maryland. The experimental results show that the early SOH prediction of Li-ion batteries can be improved by at least one order of magnitude on both the NASA and CALCE battery datasets when using the method proposed in this paper.

Data-driven forecasting framework for daily reservoir inflow time series considering the flood peaks based on multi-head attention mechanism

We propose, in this paper, a framework for time series and nonlinear system modeling, called the basis function matrix-based flexible coefficient autoregressive (BFM-FCAR) model. It has very flexible nonlinear structure. We show that many famous nonlinear time series models can be derived under this framework by choosing the proper basis function matrices. Some probabilistic properties (the conditions of geometrical ergodicity) of the BFM-FCAR model are investigated. Taking advantage of the model structure, we present an efficient parameter estimation algorithm for the proposed framework by using the variable projection method. Finally, we show how new models are generated from the proposed framework.

This paper proposes a multivariate distance nonlinear causality test (MDNC) using the partial distance correlation in a time series framework. Partial distance correlation as an extension of the Brownian distance correlation calculates the distance correlation between random vectors X and Y controlling for a random vector Z. Our test can detect nonlinear lagged relationships between time series, and when integrated with machine learning methods it can improve the forecasting power. We apply our method as a feature selection procedure and combine it with the support vector machine and random forests algorithms to study the forecast of the main energy financial time series (oil, coal, and natural gas futures). It shows substantial improvement in forecasting the fuel energy time series in comparison to the classical Granger causality method in time series.

This paper proposes a multivariate distance nonlinear causality test (MDNC) using the partial distance correlation in a time series framework. Partial distance correlation as an extension of the Brownian distance correlation calculates the distance correlation between random vectors X and Y controlling for a random vector Z. Our test can detect nonlinear lagged relationships between time series, and when integrated with machine learning methods it can improve the forecasting power. We apply our method as a feature selection procedure and combine it with the support vector machine and random forests algorithms to study the forecast of the main energy financial time series (oil, coal, and natural gas futures). It shows substantial improvement in forecasting the fuel energy time series in comparison to the classical Granger causality method in time series.

Die Konzepte der Marktintegration (MI) und Zwischenmarktpreisdynamik im internatio-nalen Handel, Warenmärkten und Industrieorganisationsgebieten stehen in direkter Ver-bindung zu Marktleistungsfähigkeit, Wettbewerbsfähigkeit und verschiedenen politischen Strategien. Demzufolge haben die Mess- und Prüfprobleme in der MI Analyse im Laufe der Jahre beträchtliche Aufmerksamkeit erhalten. Bedingt durch die Breite des Konzeptes ist es zur Einführung und die Entwicklung von divergierenden Mess-Techniken gekom-men. In dieser Studie werden die Komplexitäten der Marktintegration und der Preisübertra-gungsdynamik innerhalb von Zwischenmarktgleichgewicht-Bedingungen bewertet. Dabei sind Quellen und Implikationen von spezifischen Defiziten der gegenwärtigen nichtlinea-ren Marktintegrationsmodelle ausgehoben worden. Die Studie hat das Markov-Switching Gleichgewicht-Modell als eine vereinigende Ansicht für die umfassende Marktintegrati-onsanalyse vorgeschlagen und bewiesen. Die Flexibilität des Markovschen Rahmens wird innerhalb der Schwelle (TAR) Formulierungen verwendet, um sowohl Zwischen-markt-Gleichgewichtprozesse als auch Ergebnisse abzuleiten. In Anlehnung an die Mo-dellierungstechnik der eingeordneten Autoregression haben wir gezeigt, dass die durch Transaktionskosten auferlegten Komplikationen mit stichprobenaufspaltenden Techniken beseitigt werden können. Daher haben wir durch eine zusammengefügte Übung in der These demonstriert, dass die Flexibilität der Markovsche Formulierungen ihnen erlaubt, sowohl Anpassungstriebkräfte, die die Preisübertragungsökonometrie unterstützen als auch Gleichgewicht-Bedingungen, die die Parität-gebundenes Modell (PBM) treiben, zu behandeln.

This Chapter, investigates the relationship between gender parity in tertiary levels of education and income inequality in India, using time series econometric methodology. The Chapter attempts to explore how the gender differential in tertiary educational achievements, impacts income inequality in India over the period of 1976–2014, in a time series framework. The study finds that the situation of returns to higher education is highly unequal and it is crucial in impacting income inequality in India. The policy option suggested is to structure for a differential fee in the higher education, based on the household incomes. Scholarships can be awarded to girl students with high degrees of innate abilities. However, the feasibility of design of such programmes is a political concern and beyond the realm of academic discussion.

In present day scenario statistical (time series) and physical (NWP) models are utilized for wind power forecasting and many of them are using neural networks to obtain greater accuracy of wind power prediction at final stage. In a time series framework, forecasting is categorised into two ways single step ahead and multi-step ahead. In this paper an advanced time-series model for multi-step ahead wind power prediction based on artificial intelligence techniques is presented. This method requires an input of past measurements for prediction & input is settled on the basis of statistical tool called Auto Correlation Function (ACF). Genetic Algorithms based Neural Network (GANN) and Feed Forward Neural Network (FFNN) trained by Levenberg-Marquardt (LM) training algorithm are employed. Mean absolute error (MAE) and mean absolute percentage error (MAPE) are considered as the performance metric and both models are also compared with persistence model. The data of wind power has been collected from Ontario Electricity Market for the year 2009–12 and tested for one year up to 12 multi-steps ahead forecasting. It has been observed that GANN gives better performance as compared to FFNN.

We apply recent time series techniques to Mbaku's Cameroon data to investigate the effects of foreign aid on economic growth. In contrast to Mbaku, we find some evidence to support the foreign aid led growth hypotheses in Cameroon. Our results illustrate the importance of testing hypotheses in an appropriate time series framework and the difficulties of time-series modelling with an insufficient span of data.

Keynote Lectures.- From Clinical Guidelines to Decision Support.- Artificial Intelligence for Building Learning Health Care Organizations.- Timing Is Everything: Temporal Reasoning and Temporal Data Maintenance in Medicine.- Machine Learning for Data Mining in Medicine.- Guidelines and Protocols.- Guidelines-Based Workflow Systems.- Enhancing Clinical Practice Guideline Compliance by Involving Physicians in the Decision Process.- Application of Therapeutic Protocols: A Tool to Manage Medical Knowledge.- Decision Support Systems, Knowledge-Based Systems, Cooperative Systems.- From Description to Decision: Towards a Decision Support Training System for MR Radiology of the Brain.- Internet-Based Decision-Support Server for Acute Abdominal Pain.- Multi-modal Reasoning in Diabetic Patient Management.- Experiences with Case-Based Reasoning Methods and Prototypes for Medical Knowledge-Based Systems.- Exploting Social Reasoning of Open Multi-agent Systems to Enahnce Cooperation in Hospitals.- Influence Diagrams for Neonatal Jaundice Management.- Electronic Drug Prescribing and Administration - Bedside Medical Decision Making.- Neonatal Ventilation Tutor (VIE-NVT), a Teaching Program for the Mechanical Ventilation of Newborn Infants.- A Life-Cycle Based Authorisation Expert Database System.- A Decision-Support System for the Identification, Staging, and Functional Evaluation of Liver Diseases (HEPASCORE).- Model-Based Systems.- A Model-Based Approach for Learning to Identify Cardiac Arrhythmias.- A Model-Based System for Pacemaker Reprogramming.- Integrating Deep Biomedical Models into Medical Decision Support Systems: An Interval Constraint Approach.- Neural Networks, Causal Probabilistic Networks.- A Decision Theoretic Approach to Empirical Treatment of Bacteraemia Originating from the Urinary Tract.- An ECG Ischemic Detection System Based on Self-Organizing Maps and a Sigmoid Function Pre-processing Stage.- Neural Network Recognition of Otoneurological Vertigo Diseases with Comparison of Some Other Classification Methods.- A Comparison of Linear and Non-linear Classifiers for the Detection of Coronary Artery Disease in Stress-ECG.- The Case-Based Neural Network Model and Its Use in Medical Expert Systems.- Knowledge Representation.- A Medical Ontology Library That Integrates the UMLS Metathesaurus(TM).- The Use of the UMLS Knowledge Sources for the Design of a Domain Specific Ontology: A Practical Experience in Blood Transfusion.- Representing Knowledge Levels in Clinical Guidelines.- Temporal Reasoning.- Intelligent Analysis of Clinical Time Series by Combining Structural Filtering and Temporal Abstractions.- Knowledge-Based Event Detection in Complex Time Series Data.- Abstracting Steady Qualitative Descriptions over Time from Noisy, High-Frequency Data.- Visualization Techniques for Time-Oriented, Skeletal Plans in Medical Therapy Planning.- Visualizing Temporal Clinical Data on the WWW.- Machine Learning.- Machine Learning in Stepwise Diagnostic Process.- Refinement of Neuro-psychological Tests for Dementia Screening in a Cross Cultural Population Using Machine Learning.- The Analysis of Head Injury Data Using Decision Tree Techniques.- Machine Learning for Survival Analysis: A Case Study on Recurrence of Prostate Cancer.- ICU Patient State Characterization Using Machine Learning in a Time Series Framework.- Diagnostic Rules of Increased Reliability for Critical Medical Applications.- Machine Learning Inspired Approaches to Combine Standard Medical Measures at an Intensive Care Unit?.- A Screening Technique for Prostate Cancer by Hair Chemical Analysis and Artificial Intelligence.- Natural Language Processing.- A Conversational Model for Health Promotion on the World Wide Web.- Types of Knowledge Required to Personalise Smoking Cessation Letters.- Small Is Beautiful - Compact Semantics for Medical Language Processing.- Speech Driven Natural Language Understanding for Hands-Busy Recording of Clinical Information.- Automatic Acquisition of Morphological Knowledge for Medical Language Processing.- Image Processing and Computer Aided Design.- A Multi-agent System for MRI Brain Segmentation.- Modelling Blood Vessels of the Eye with Parametric L-Systems Using Evolutionary Algorithms.- Animating Medical and Safety Knowledge.- Active Shape Models for Customised Prosthesis Design.

In present day scenario statistical (time series) and physical (NWP) models are utilized for wind power forecasting and many of them are using neural networks to obtain greater accuracy of wind power prediction at final stage. In a time series framework, forecasting is categorised into two ways single step ahead and multi-step ahead. In this paper an advanced time-series model for multi-step ahead wind power prediction based on artificial intelligence techniques is presented. This method requires an input of past measurements for prediction & input is settled on the basis of statistical tool called Auto Correlation Function (ACF). Genetic Algorithms based Neural Network (GANN) and Feed Forward Neural Network (FFNN) trained by Levenberg-Marquardt (LM) training algorithm are employed. Mean absolute error (MAE) and mean absolute percentage error (MAPE) are considered as the performance metric and both models are also compared with persistence model. The data of wind power has been collected from Ontario Electricity Market for the year 2009-12 and tested for one year up to 12 multi-steps ahead forecasting. It has been observed that GANN gives better performance as compared to FFNN.

Control charts contribute to the monitoring and improvement of process quality by helping to separate out special cause variation from common cause variation. By common cause variation we mean the usual variation in an in-control process. Special causes can be thought of as disturbances, possibly transitory, impacting a process that is in a state of statistical control. However, there is no clear place in this scheme of special causes and common causes for systematic non-iid variation, such as trend, seasonal, autoregression variation, and intervention effects from efforts to improve the proess. When systematic non-iid variation is present, time series modeling and fitting can fill in this picture. In the time series framework, observations influenced by special causes can be treated as outliers from the currently-entertained time-series model and can be detected by outlier detection methods. We discuss three data sets that illustrate how this can be done in order to make control charts more effective. We show also how a standard control-chart supplement called "pattern analysis" can be useful in time-series work.

CSmoothing allows an analyst to use the so-called Controlled Smoothing technique to estimate trends in a time series framework. In this Web-tool (Shiny), the analyst may apply the methodology to at most 3 mortality time series simultaneously, as well as to other kind of time series individually. Likewise, this smoothing approach allows the analyst to establish one, two or three segments in order to take into account possible changes in variance regimes. For estimating trends it uses different amounts of smoothness, both globally for the total data set and through some partial indices for each selected segment. It is also possible to endogenously fix the points where the segments start and end (the cutoff points) with continuous joints. Additionally, intervals of different standard deviations for their respective trends are given. Particular emphasis is placed on a big data set of log mortality rates, log(qx), taken from period life tables of the Human Mortality Database (HMD) (University of California Berkeley (USA) and and Max Planck Institute for Demographic Research (Germany)), 2021). In all cases, dynamic graphs and several statistics related to the Controlled Smoothing technique are illustrated.