Research team
Expertise
Development and study of advanced data mining and machine learning methods. In particular, we investigate: (i) new methods to efficiently discover interesting patterns in sequential data; (ii) new methods to detect contextual anomalies in heterogeneous sequential data; (iii) and new methods for multi-label classification in extremely large datasets. In addition, we investigate applications of these methods in areas such as the monitoring of wind farms and anomaly detection in an Industrial Internet of Things context.
Interpretable rule-based recommender systems.
Abstract
Recommender systems help users identify the most relevant items from a huge catalogue. In recent independent evaluation studies of recommender systems, baseline association rule models are competitive with more complex state-of-the-art methods. Moreover, rule-based recommender algorithms have several exciting properties, such as the potential to be interpretable, the ability to identify local patterns and the support of context-aware predictions. First, we survey various existing recommendation algorithms with different biases and prediction strategies and evaluate them independently. Besides accuracy, we evaluate coverage and diversity and analyse the structure of the resulting rule models, which are essential towards understanding interpretability. Second, we propose to gap the bridge between recommender systems and recent multi-label classification based on learning an optimal set of rules w.r.t. to a custom loss function. We study if a decision-theoretic framework can guarantee the identification of the optimal rules for recommender systems under a loss function combining accuracy, complexity and diversity. We account for characteristics unique to recommender datasets, such as skewed distribution, implicit feedback and scale. Finally, we adopt new rule-based algorithms that are interpretable and more accurate. We apply them for healthcare recommendations to improve intensive care unit monitoring and online bandit learning for large-scale websites for e-commerce and news.Researcher(s)
- Promoter: Goethals Bart
- Fellow: Feremans Len
Research team(s)
Project type(s)
- Research Project
Interpretable Qualitative Evaluation for Online Recommender Systems.
Abstract
Individuals often rely on recommendations provided by others in making routine, daily decisions. Algorithms, mimicking this behaviour, are vital to the success of e-commerce services. However, a remaining open question is why algorithms make these recommendations. This is problematic given that, the most accurate machine learning algorithms are black-box models, and we have a dynamic environment were possibly multiple models are deployed and periodically re-trained. Since any organisation requires human oversight and decision-making, there is a need for insight into user behaviour and interactions with recommendations made by black-box machine learning algorithms. Traditionally, two recommender systems are compared based on a single metric, such as click-through-rate after an A/B test. We will assess the performance of online recommender systems qualitatively by uncovering patterns that are characteristic for the differences in targeted users and items. We propose to adopt interpretable machine learning, where the goal is to produce explanations that can be used to guide processes of human understanding and decisions. We propose to mine interpretable association rules and generate, possibly grouped, counterfactual explanations why recommender system A performs better (or worse) than recommender system B.Researcher(s)
- Promoter: Goethals Bart
- Fellow: Feremans Len
Research team(s)
Project type(s)
- Research Project