Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Towards Explainable Sequential Learning

Giacomo Bergami, Emma Packer, Kirsty Scott, Silvia Del Din

TL;DR

This work addresses the gap between numerical and event-based temporal data analysis by introducing a hybrid explainable sequential learning pipeline, EMeriTAte+DF, that discretizes multivariate numerical data into polyadic logs and couples declarative specifications with white-box learning to produce human-interpretable models. It extends DECLARE-based specification mining to polyadic, concurrent events and enriches clauses with dataful predicates (Catch22 features) to capture nuanced temporal dynamics. The approach yields strong classification performance across diverse datasets and provides scalable, explainable representations through a priori indexing/mining and ad hoc refinements, with improved runtime efficiency and explicit formalism. The contribution offers a versatile, explainable framework for real-world temporal classification that explicitly models cross-dimensional correlations and concurrent event structures, enabling practical deployment in domains like healthcare and mobility analytics.

Abstract

This paper offers a hybrid explainable temporal data processing pipeline, DataFul Explainable MultivariatE coRrelatIonal Temporal Artificial inTElligence (EMeriTAte+DF), bridging numerical-driven temporal data classification with an event-based one through verified artificial intelligence principles, enabling human-explainable results. This was possible through a preliminary a posteriori explainable phase describing the numerical input data in terms of concurrent constituents with numerical payloads. This further required extending the event-based literature to design specification mining algorithms supporting concurrent constituents. Our previous and current solutions outperform state-of-the-art solutions for multivariate time series classifications, thus showcasing the effectiveness of the proposed methodology.

Towards Explainable Sequential Learning

TL;DR

This work addresses the gap between numerical and event-based temporal data analysis by introducing a hybrid explainable sequential learning pipeline, EMeriTAte+DF, that discretizes multivariate numerical data into polyadic logs and couples declarative specifications with white-box learning to produce human-interpretable models. It extends DECLARE-based specification mining to polyadic, concurrent events and enriches clauses with dataful predicates (Catch22 features) to capture nuanced temporal dynamics. The approach yields strong classification performance across diverse datasets and provides scalable, explainable representations through a priori indexing/mining and ad hoc refinements, with improved runtime efficiency and explicit formalism. The contribution offers a versatile, explainable framework for real-world temporal classification that explicitly models cross-dimensional correlations and concurrent event structures, enabling practical deployment in domains like healthcare and mobility analytics.

Abstract

This paper offers a hybrid explainable temporal data processing pipeline, DataFul Explainable MultivariatE coRrelatIonal Temporal Artificial inTElligence (EMeriTAte+DF), bridging numerical-driven temporal data classification with an event-based one through verified artificial intelligence principles, enabling human-explainable results. This was possible through a preliminary a posteriori explainable phase describing the numerical input data in terms of concurrent constituents with numerical payloads. This further required extending the event-based literature to design specification mining algorithms supporting concurrent constituents. Our previous and current solutions outperform state-of-the-art solutions for multivariate time series classifications, thus showcasing the effectiveness of the proposed methodology.

Paper Structure

This paper contains 25 sections, 2 theorems, 1 equation, 4 figures, 2 tables, 5 algorithms.

Table of Contents

  1. Introduction
  2. Related Works
  3. Hybridly Explainable Artificial Intelligence
  4. A Priori Explanability
  5. Numerical Data.
  6. Event Data.
  7. Ad Hoc Explanability
  8. Numerical Data.
  9. Event Data.
  10. Problem Statement
  11. Polyadic Logs
  12. Poly-DECLARE
  13. A Priori Explainability
  14. Loading and Indexing.
  15. Indexed Mining.
  16. ...and 10 more sections

Key Result

lemma thmcounterlemma

Given a collection of $N$ of $d$ dimensions with maximum length $t$, the time complexity is in $O(Ndt)$.

Figures (4)

  • Figure 1: Declarative Languages for a priori explanability over (a) or (b) non-polyadic logs.
  • Figure 2: Fine-Grained Mining+Learning times for the training phase of both and .
  • Figure 3: Comparing the cumulative training times.
  • Figure 4: Training Results over the datasets of interest. Macro metrics are used for datasets containing more than 2 classes. Numbers in blue (red) remark the best (worst) results.

Theorems & Definitions (4)

  • lemma thmcounterlemma: Indexing and Loading Time Complexity
  • proof
  • lemma thmcounterlemma
  • proof