Objectives

This is a protocol for a Cochrane Review (prognosis). The objectives are as follows:

To identify and evaluate AI‐based prognostic models, in development and external validation studies, which predict impending episodes of ventricular tachycardia from continuous surface electrocardiogram (ECG) recordings in individuals at elevated risk for ventricular arrhythmias due to any relevant underlying condition.

Digital twins are virtual replicas of their physical counterparts, providing real-time monitoring and decision-making capabilities. By integrating forecasting-based methods, the potential of digital twins can be augmented significantly, enabling them to execute advanced predictive tasks. However, with digital twins typically involving a human-in-the-loop, the need for explainability becomes crucial for understanding how and why a forecast was made. To effectively integrate explainability methods, forecasting methods, and digital twins, it is essential to define the relations between these components in a structured manner. In this work, we address this issue by providing a meta-model for the integration of explainable forecasting methods with digital twins. We evaluate our meta-model in the context of a smart building digital twin with multiple forecasting and explainability methods. The evaluation demonstrates the inherent trade-off between providing explanations and generating accurate forecasts in this context.

Melanoma is the most common form of skin cancer, responsible for thousands of deaths annually. Novel therapies have been developed, but metastases are still a common problem, increasing the mortality rate and decreasing the quality of life of those who experience them. As traditional machine learning models for metastasis prediction have been limited to the use of a single modality, in this study we aim to explore and compare different unimodal and multimodal machine learning models to predict the onset of metastasis in melanoma patients to help clinicians focus their attention on patients at a higher risk of developing metastasis, increasing the likelihood of an earlier diagnosis.

We use a patient cohort derived from an Electronic Health Record, and we consider various modalities of data, including static, time series, and clinical text. We formulate the problem and propose a multimodal ML workflow for predicting the onset of metastasis in melanoma patients. We evaluate the performance of the workflow based on various classification metrics and statistical significance. The experimental findings suggest that multimodal models outperform the unimodal ones, demonstrating the potential of multimodal ML to predict the onset of metastasis.

Digital twin systems can benefit from the integration of artificial intelligence (AI) algorithms for providing for example some predictive capabilities or supporting internal decision-making. As AI algorithms are often opaque, it becomes necessary to explain their decisions to a human operator working with the digital twin. In this study, we investigate the integration of explainable AI techniques with digital twins, which we termed XAI-DT system. We define the concept of XAI-DT system and provide a use case in smart buildings, where explainable AI is used to forecast CO₂ concentration. Further, we present a core architectural model for our digital twin, outlining its interaction with the smart building and its internal processing. We evaluate five AI algorithms and compare their explainability for the operator and the entire digital twin model based on standard explainability properties from the literature.

Energy management requires reliable tools to support decisions aimed at optimising consumption. Advances in data-driven models provide techniques like Non-Intrusive Load Monitoring (NILM), which estimates the energy demand of appliances from total consumption. Common single-target NILM approaches perform energy disaggregation by using separate learned models for each device. However, the use of single-target systems in real scenarios is computationally expensive and can obscure the interpretation of the resulting feedback. This study assesses a conditioned deep neural network built upon a Fully Convolutional Denoising AutoEncoder (FCNdAE) as multi-target NILM model. The network performs multiple disaggregations using a conditioning input that allows the specification of the target appliance. Experiments compare this approach with several single-target and multi-target models using public residential data from households and non-residential data from a hospital facility. Results show that the multi-target FCNdAE model enhances the disaggregation accuracy compared to previous models, particularly in non-residential data, and improves computational efficiency by reducing the number of trainable weights below 2 million and inference time below 0.25 s for several sequence lengths. Furthermore, the conditioning input helps the user to interpret the model and gain insight into its internal behaviour when predicting the energy demand of different appliances.

Big time series are increasingly available from an ever wider range of IoT-enabled sensors deployed in various environments. Significant insights can be gained by mining temporal patterns from these time series. Temporal pattern mining (TPM) extends traditional pattern mining by adding event time intervals into extracted patterns, making them more expressive at the expense of increased time and space complexities. Besides frequent temporal patterns (FTPs), which occur frequently in the entire dataset, another useful type of temporal patterns are so-called rare temporal patterns (RTPs), which appear rarely but with high confidence. Mining rare temporal patterns yields additional challenges. For FTP mining, the temporal information and complex relations between events already create an exponential search space. For RTP mining, the support measure is set very low, leading to a further combinatorial explosion and potentially producing too many uninteresting patterns. Thus, there is a need for a better approach to mine frequent and rare temporal patterns. This paper presents our Generalized Temporal Pattern Mining from Time Series (GTPMfTS) approach that can mine both types of patterns, with the following specific contributions: (1) The end-to-end GTPMfTS process taking time series as input and producing frequent/rare temporal patterns as output. (2) The efficient Generalized Temporal Pattern Mining (GTPM) algorithm mines frequent and rare temporal patterns using efficient data structures for fast retrieval of events and patterns during the mining process, and employs effective pruning techniques for significantly faster mining. (3) An approximate version of GTPM that uses mutual information, a measure of data correlation, to prune unpromising time series from the search space. (4) An extensive experimental evaluation of GTPM for rare temporal pattern mining (RTPM) and frequent temporal pattern mining (FTPM), showing that RTPM and FTPM significantly outperform the baselines on runtime and memory consumption, and can scale to big datasets. The approximate RTPM is up to one order of magnitude, and the approximate FTPM is up to two orders of magnitude, faster than the baselines, while retaining high accuracy.

There is an increasing need to include explainability on the machine learning (ML) models. Among the various approaches, counterfactual (CF) explanations allow the design of what-if scenarios and the interactive exploration of ML model behavior on sensitive decision-making domains. However, the generation of CF for tabular and time-series data requires technical skills that are not always available to the end-users of ML-powered systems. Therefore, we propose a modular web-based tool to easily generate, visualize, and interact with CF on any tabular or time-series dataset. The EXTREMUM platform provides access to state-of-the-art CF algorithms, where users can train ML models and explore CF on their tabular or time-series datasets with an intuitive user interface. The project is instantiated on two tabular datasets within healthcare and five time-series datasets with various domains. The open-source repository lets ML researchers adapt the existing ML tool to new application domains: https://gitea.dsv.su.se/DataScienceGroup/EXTREMUM-demo.

Designing affect-based personalized technology involves dealing with large datasets. Machine learning (ML) algorithms are employed to predict affect and similar human factors from in-game metrics, behavioral patterns, or physiological responses. The classification performance is usually presented as a global point estimate without providing user-specific interpretations. This approach is incompatible with effective personalization in games because it disregards the variability of body responses between players. This paper proposes a methodology to classify subjective human factors from large multimodal data. A public VR dataset (CEAP-360VR) was used to extensively compare three ML classifiers and five feature importance techniques. The produced models could reduce the original feature space by 82% (from 113 to 20 features) without compromising predictive performance (F1 score). A random forest (RF) using forward sequential feature selection (fSFS) yielded the best prediction of binary valence (F1=0.761) and arousal (F1=0.748). Finally, feature importance rankings are discussed with emphasis on global and user-specific patterns that may improve affect recognition. The proposed methodology is envisioned to help game designers and researchers create customized user-centric games and VR experiences inferring possible explanations from multimodal datasets.

CounterFair is a group counterfactual search algorithm that detects and minimizes biases among sensitive groups and identifies relevant subgroups inside these sensitive groups based on shared counterfactual instances. We investigate the latter capability, analyzing the found subgroups from the perspective of fairness based on counterfactual reasoning, in order to evaluate whether they present different biases with respect to each other and to the sensitive feature groups they belong to. We perform these measurements on the subgroups extracted by CounterFair over six binary classification datasets, providing figures and their respective analysis on the presence of bias.

For the time series forecasting task, several state-of-the-art algorithms employ moving-average decomposition for improved accuracy. However, the potential of decomposition techniques to enhance time series forecasting methods has not been explored in detail. In this work, we comprehensively investigate the use of decomposition methods for the forecasting task, comparing different decomposition techniques and their effect on forecasting accuracy, as well as the possibility of providing model-agnostic interpretability. We rework recent forecasting models to be compatible with any decomposition technique and experimentally evaluate their effectiveness in different forecasting setups. We further propose and assess a model-agnostic framework using decomposition for interpretability. Our results show that decomposition can improve forecasting accuracy, especially for the proposed decomposition-adapted models. Additionally, we demonstrate that the architectural choices of existing forecasting models can be improved by using different decomposition blocks internally. We found that decomposition techniques must be configured with a low number of components to provide model-agnostic interpretability. Our work concludes that decomposition can enhance time series forecasting algorithms, improving both their performance and interpretability.