This book constitutes the refereed proceedings of the 4th ECML PKDD Workshop on Advanced Analytics and Learning on Temporal Data, AALTD 2019, held in Ghent, Belgium, in September 2020.
The 15 full papers presented in this book were carefully reviewed and selected from 29 submissions. The selected papers are devoted to topics such as Temporal Data Clustering; Classification of Univariate and Multivariate Time Series; Early Classification of Temporal Data; Deep Learning and Learning Representations for Temporal Data; Modeling Temporal Dependencies; Advanced Forecasting and Prediction Models; Space-Temporal Statistical Analysis; Functional Data Analysis Methods; Temporal Data Streams; Interpretable Time-Series Analysis Methods; Dimensionality Reduction, Sparsity, Algorithmic Complexity and Big Data Challenge; and Bio-Informatics, Medical, Energy Consumption, Temporal Data.
Author(s): Vincent Lemaire, Simon Malinowski, Anthony Bagnall, Thomas Guyet, Romain Tavenard, Georgiana Ifrim
Series: Lecture Notes in Computer Science, 12588
Publisher: Springer
Year: 2021
Language: English
Pages: 233
City: Cham
Preface
Workshop Description
Organization
Contents
Oral Presentation
On the Usage and Performance of the Hierarchical Vote Collective of Transformation-Based Ensembles Version 1.0 (HIVE-COTE v1.0)
1 Introduction
2 HIVE-COTE 1.0 Design
2.1 Ensemble Structure
2.2 Time Series Forest (TSF)
2.3 Random Interval Spectral Ensemble (RISE)
2.4 Bag of SFA Symbols (BOSS)
2.5 Shapelet Transform Classifier (STC)
3 HIVE-COTE 1.0 Usability
3.1 Java Implementation of HIVE-COTE 1.0 in tsml
3.2 Python Implementation of HIVE-COTE 1.0 in sktime
4 Performance
5 Conclusions
References
Ordinal Versus Nominal Time Series Classification
1 Introduction
2 Background
2.1 Time Series Shapelets
2.2 Ordinal Classification
3 Experimental Results and Discussion
3.1 TSOC Datasets
3.2 Experimental Settings
3.3 Results
3.4 Comparison Against the State-of-the-Art Algorithms in TSC
4 Conclusions
References
Generalized Chronicles for Temporal Sequence Classification
1 Introduction
2 Related Works
3 Discriminant Chronicle Mining
4 Generalized Discriminant Chronicles (GDC)
4.1 Taking Decisions with Generalized Discriminant Chronicles
4.2 Learning Generalized Discriminant Chronicles Classifiers
5 Examples of GDC Instances
6 Experiments
6.1 Experimental Setup
6.2 Results
7 Conclusion and Perspectives
References
Demand Forecasting in the Presence of Privileged Information
1 Introduction
2 Related Work
3 A Privileged Information-Aware Neural Network
3.1 Problem Statement
3.2 Architecture Overview
3.3 Architecture Details
3.4 Learning Process
4 Experimental Setup
5 Experimental Results
5.1 Capturing the Effects of Privileged Information
5.2 Comparison to Existing Approaches for Demand Forecasting
6 Conclusions and Future Work
References
GANNSTER: Graph-Augmented Neural Network Spatio-Temporal Reasoner for Traffic Forecasting
1 Introduction
2 Related Work
2.1 Traffic Forecasting
2.2 Graph Neural Networks
2.3 Graph Neural Networks for Traffic Forecasting
3 GANNSTER
3.1 Road Graph
3.2 Definitions
3.3 GANNSTER Model
4 Experimental Evaluation
4.1 MUSTARD-S
4.2 Experimental Settings
5 Results and Discussion
6 Conclusion and Future Work
References
A Model-Agnostic Approach to Quantifying the Informativeness of Explanation Methods for Time Series Classification
1 Introduction
2 Related Work
2.1 Time Series Classification
2.2 Explanation in Time Series Classification
2.3 Explanation in Other Machine Learning Domains
3 Research Methods
3.1 Explanation-Driven Perturbation
3.2 Method 1: Evaluating a Single Explanation Method
3.3 Method 2: Comparing Multiple Explanation Methods
3.4 Informativeness of an Explanation: An Evaluation Measure
4 Experiments
4.1 Experiment 1: Evaluation of a Single Explanation Method
4.2 Comparison of Multiple Explanation Methods
4.3 Sanity Checks for Experiment Results
5 Discussion
6 Conclusion
References
Poster Presentation
Temporal Exceptional Model Mining Using Dynamic Bayesian Networks
1 Introduction
2 Motivating Example: The Business Process Intelligence Challenge
3 Temporal Exceptional Model Mining
3.1 Temporal Targets
3.2 Subgroups
3.3 Problem Statement
4 Exceptional Dynamic Bayesian Networks
4.1 Dynamic Bayesian Networks
4.2 Distance Function
4.3 Scoring Function
4.4 Exceptional Subgroups
4.5 Distribution of False Discoveries
4.6 Subgroup Search
4.7 Exceptionality Test
5 Experiments with Simulated Data
5.1 Data Generating Procedure
5.2 Evaluation
5.3 Results
5.4 Impact of (dis)similar Models on Prediction
6 Data of Funding Applications
6.1 Data
6.2 Discovered Subgroups
6.3 Comparison to Previous Analyses
6.4 Subgroup Differences
7 Related Work
8 Conclusions
References
``J'veux du Soleil'' Towards a Decade of Solar Irradiation Data (La Réunion Island, SW Indian Ocean)
1 Introduction
2 Data Acquisition
2.1 Measuring Irradiation and Meteorological Parameters
3 Available Data
4 Valuable Applications
5 Conclusion
A Technical Setup
References
Visual Analytics for Extracting Trends from Spatio-temporal Data
1 Introduction
2 Rationale and Related Work
2.1 Rationale
2.2 Related Work
3 Methods
3.1 Cluster Analysis
3.2 Temporal Fingerprinting Through Circular Heat Maps
3.3 Spatio-temporal Comparison Through Circular Heat Map Subtraction
3.4 Temporal Behaviour Characterisation Through Label Maps
4 Case Study on Brussels Traffic
4.1 Data
4.2 Unravelling Volume Patterns of Brussels Traffic
4.3 Insightful Blueprints of Brussels Traffic
5 Conclusion and Future Work
References
Layered Integration Approach for Multi-view Analysis of Temporal Data
1 Introduction
2 Related Work and Rationale
2.1 Challenges Related to Real-World Datasets
2.2 Multi-view Learning
3 Use Case Context and Ambition
4 Methods and Proposed Approach
4.1 Clustering Analysis
4.2 Kernel Density Estimation (KDE)
4.3 Hypercube Binning Approach
4.4 Layered Multi-view Analysis: General Approach
4.5 Layered Multi-view Analysis: Instantiated in the Use Case
5 Dataset and Implementation
5.1 Data Preprocessing
5.2 Implementation and Availability
6 Results and Discussion
6.1 Individual Analysis Layer: Operating Mode Characterisation
6.2 Mediation Layer: Performance Profiling
6.3 Integration Layer: Fleet-Wide Performance Labelling
7 Conclusion and Future Work
References
Real-Time Outlier Detection in Time Series Data of Water Sensors
1 Introduction
2 Data Overview
3 Experiment Setup
3.1 Outlier Detection Pipeline
3.2 Synthetic Evaluation
4 Modelling and Hyper-Parameter Tuning
4.1 Regression-Based Models
4.2 Neural Network-Based Approaches
4.3 Direct Classification Model: Isolation Forests (IF)
5 Results
5.1 Illustrative Examples: Univariate Results
5.2 Illustrative Example: Multivariate Results QR-MLP
5.3 Comparison of Univariate and Multivariate Modelling Techniques
5.4 Comparison of Multivariate Modelling Techniques
5.5 Practical Impact
6 Conclusion and Future Work
References
Lightweight Temporal Self-attention for Classifying Satellite Images Time Series
1 Introduction
2 Method
2.1 Multi-headed Self-attention
2.2 Lightweight Attention
2.3 Spatio-Temporal Classifier
3 Numerical Experiment
3.1 Dataset
3.2 Metric and Protocol
3.3 Evaluated Methods
3.4 Analysis
3.5 Ablation Study and Robustness Assessment
3.6 Computational Complexity
4 Conclusion
References
Creating and Characterising Electricity Load Profiles of Residential Buildings
1 Introduction
2 Related Work
3 Smart Meter Characterisation and Classification Problem
4 Methodology
4.1 Time Series Clustering
4.2 Survey Classification
5 Experimental Results
5.1 Feature Vectors
5.2 Behaviour Clusters
5.3 Cluster Classification
5.4 Feature Importance
6 Conclusions
A Survey Questions
References
Trust Assessment on Streaming Data: A Real Time Predictive Approach
1 Introduction
2 Related Work
3 Data Trustworthiness Online Model: DTOM
3.1 Problem Statement
3.2 Design and Implementation
3.3 Online Ensemble Regression
4 Experimentation
4.1 Experimental Dataset
4.2 Evaluation
4.3 Results
5 Conclusion
References
A Feature Selection Method for Multi-dimension Time-Series Data
1 Introduction
2 Feature Selection
2.1 Correlation Based Feature Selection Using Mutual Information
2.2 Feature Selection for Time-Series Data
3 CFS for Time-Series Data
4 Evaluation
4.1 Data Sets
4.2 Merit Score Evaluation
4.3 Feature Subset Selection
5 Conclusions and Future Work
References
Author Index
