Shapley Value Research Articles

Machine learning algorithms for prediction of measles one vaccination dropout among 12–23 months children in Ethiopia

IntroductionDespite the availability of a safe and effective measles vaccine in Ethiopia, the country has experienced recurrent and significant measles outbreaks, with a nearly fivefold increase in confirmed cases from...

Understanding Polymers Through Transfer Learning and Explainable AI

In this work we study the use of artificial intelligence models, particularly focusing on transfer learning and interpretability, to predict polymer properties. Given the challenges imposed by data scarcity in polymer science, transfer learning offers a promising solution by using learnt features of models pre-trained on other datasets. We conducted a comparative analysis of direct modelling and transfer learning-based approaches using a polyacrylates’ glass transitions dataset as a proof-of-concept study. The AI models utilized tokenized SMILES strings to represent polymer structures, with convolutional neural networks processing these representations to predict Tg. To enhance model interpretability, Shapley value analysis was employed to assess the contribution of specific chemical groups to the predictions. The results indicate that while transfer learning provides robust predictive capabilities, direct modelling on polymer-specific data offers superior performance, particularly in capturing the complex interactions influencing Tg. This work highlights the importance of model interpretability and the limitations of applying molecular-level models to polymer systems.

Heart Failure Prediction Using Machine Learning

This study explores the use of XGBoost and Shapley values for heart failure prediction, in addition to processing models such as neural network, CNN, and SVM. By comparing the performance of these different machine learning algorithms, we aim to develop a robust and accurate predictive model for identifying individuals at risk of heart failure. Our findings highlight the strengths and limitations of each approach, providing valuable insights for future research in this area. Ultimately, this research aims to improve the early detection and management of heart failure, leading to better patient outcomes and more efficient healthcare practices. Key Words: heart failure prediction, XGBoost, SHAPley values, neural network, CNN, SVM, machine learning algorithms, early detection, death events.

Investigation of the relationship between digital obesity status and alexithymia in young people aged 10-24

In today's rapidly digitalizing world, concepts such as digital obesity and alexithymia are becoming increasingly important. The aim of this study is to examine the concepts of digital obesity and alexithymia to understand the effects of these phenomena and prevention behaviors. In this way, it is aimed to raise awareness about the effects of digital technologies on health and contribute to individuals developing healthier digital habits. This study carried out a descriptive correlational study with individuals between the ages of 10-24 (n: 391) in schools affiliated with the Ağrı Provincial Directorate of National Education and Ağrı İbrahim Çeçen University between May 2023 and July 2023. The SPSS-22 package program and R programming language 4.1.3 were used in the study. It was found that the level of alexithymia increased as the digital obesity status of young people increased (t = 7.978; p = 0.001). Furthermore, it was determined that the Elastic Net Regression method is the best model that can be used to predict the alexithymia level. According to the Shapley value, the most important variable in predicting alexithymia was found to be digital obesity. The study determined that as the level of digital obesity increased, alexithymia also increased. In the fight against digital obesity and alexithymia, educational programs and guidance services can be provided to encourage individuals to develop conscious digital usage habits. Keywords: young people, digital obesity, alexithymia

Better Apprenticeship Learning with LLM Explanations

As the population ages, care robots will play an increasing role in assisting caregiving by taking on repetitive or physically cumbersome activities. To effectively provide care, robotic agents must be able to meet the needs and preferences of care receivers. However, these needs and preferences may change over time, making it intractable to pre-define the way the care robot should act before deployment. Instead, the care robot should be able to learn directly from non-expert end-user demonstrations. However, prior work investigating the feasibility of learning a policy from older adult demonstrations finds that older adult demonstrators desire a better understanding of what the robot needs them to do, and how. To help demonstrators understand how to improve on suboptimal or heterogeneous demonstrations, we propose to utilize a Large Language Model to provide human-interpretable explanations of Shapley values of a policy. These explanations enable the demonstrator to understand how the policy is performing, and what changes are needed, informing their corrective demonstrations. We showcase our framework's performance in deterministic and stochastic versions of Wumpus World.

Data-Driven Analysis of Ocean Fronts’ Impact on Acoustic Propagation: Process Understanding and Machine Learning Applications, Focusing on the Kuroshio Extension Front

Ocean fronts, widespread across the global ocean, cause abrupt shifts in physical properties such as temperature, salinity, and sound speed, significantly affecting underwater acoustic communication and detection. While past research has concentrated on qualitative analysis and small-scale research on ocean front sections, a comprehensive analysis of ocean fronts’ characteristics and their impact on underwater acoustics is lacking. This study employs high-resolution reanalysis data and in situ observations to accurately identify ocean fronts, sound speed structures, and acoustic propagation features from over six hundred thousand Kuroshio Extension Front (KEF) sections. Utilizing marine big data statistics and machine learning evaluation metrics such as out-of-bag (OOB) error and Shapley values, this study quantitatively assesses the variations in sound speed structures across the KEF and their effects on acoustic propagation shifts. This study’s key findings reveal that differences in sound speed structure are significantly correlated with KEF strength, with the channel axis depth and conjugate depth increasing with front strength, while the thermocline intensity and depth excess decrease. Acoustic propagation features in the KEF environment exhibit notable seasonal variations.

Stability for Coalition Structures in Terms of the Proportional Partitional Shapley Value

AbstractMany coalitional values have been introduced in the literature on cooperative games over the last decades, and especially since 2000. The multiplicity of options suggests the convenience of testing the existence of stable coalition structures, in the sense of Hart and Kurz (1983, Econometrica), when payments are made using some of these values. We recall their concept of $$\gamma$$ γ –stability and give results for the proportional partitional Shapley value, introduced by Alonso–Meijide et al (2015, Discrete Appl. Math.), which shares the utility of any coalition proportionally to the Shapley value of the involved players in the original game.

Non-small cell lung cancer detection through knowledge distillation approach with teaching assistant.

Non-small cell lung cancer (NSCLC) exhibits a comparatively slower rate of metastasis in contrast to small cell lung cancer, contributing to approximately 85% of the global patient population. In this work, leveraging CT scan images, we deploy a knowledge distillation technique within teaching assistant (TA) and student frameworks for NSCLC classification. We employed various deep learning models, CNN, VGG19, ResNet152v2, Swin, CCT, and ViT, and assigned roles as teacher, teaching assistant and student. Evaluation underscores exceptional model performance in performance metrics achieved via cost-sensitive learning and precise hyperparameter (alpha and temperature) fine-tuning, highlighting the model's efficiency in lung cancer tumor prediction and classification. The applied TA (ResNet152) and student (CNN) models achieved 90.99% and 94.53% test accuracies, respectively, with optimal hyperparameters (alpha = 0.7 and temperature = 7). The implementation of the TA framework improves the overall performance of the student model. After obtaining Shapley values, explainable AI is applied with a partition explainer to check each class's contribution, further enhancing the transparency of the implemented deep learning techniques. Finally, a web application designed to make it user-friendly and classify lung types in recently captured images. The execution of the three-stage knowledge distillation technique proved efficient with significantly reduced trainable parameters and training time applicable for memory-constrained edge devices.

Cost-sharing in Parking Games

In this paper, we study the total displacement statistic of parking functions from the perspective of cooperative game theory. We introduce parking games, which are coalitional cost-sharing games in characteristic function form derived from the total displacement statistic. We show that parking games are supermodular cost-sharing games, indicating that cooperation is difficult (i.e., their core is empty). Next, we study their Shapley value, which formalizes a notion of "fair" cost-sharing and amounts to charging each car for its expected marginal displacement under a random arrival order. Our main contribution is a polynomial-time algorithm to compute the Shapley value of parking games, in contrast with known hardness results on computing the Shapley value of arbitrary games. The algorithm leverages the permutation-invariance of total displacement, combinatorial enumeration, and dynamic programming. We conclude with open questions around an alternative solution concept for supermodular cost-sharing games and connections to other areas in combinatorics.

Shear performance prediction for corrugated steel web girders based on machine-learning algorithms

This study aimed to predict the shear strength of corrugated steel web girders (CSWGs) by developing a new method based on four machine-learning (ML) algorithms, namely the support vector machine, artificial neural network, random forest, and XGBoost. Based on the acquired experimental and numerical data, a database containing 552 samples was constructed to train and test the ML models. A five-fold cross-validation approach was adopted during training to prevent model overfitting. A RandomizedSearchCV was used to optimize the hyperparameters of each model. The performance of the trained models was evaluated using four performance metrics, and the results revealed that the coefficients of determination (R2) of all ML models exceeded 0.97 when used on both training and validation sets, demonstrating the excellent performance of the ML models in predicting the shear strength of CSWGs. Additionally, the implemented ML models outperformed existing design codes and empirical formulae. The XGBoost model yielded the best prediction results with an R2 of 0.999, mean absolute error of 44.98 kN, root-mean-square error of 66.67 kN, and mean absolute percentage error of 2.1 %. By using the Shapley additive explanation to derive a visual, quantitative explanation of the XGBoost model, the yield strength, web thickness, and web height were identified as the most critical factors affecting the shear strength of CSWGs, and their average absolute Shapley values accounted for approximately 91.45 % of the total value. The ML models implemented in this study provide a promising new approach for pre-designing and verifying the stability of CSWGs.

Feature Importance: A Closer Look at Shapley Values and LOCO

Feature Importance: A Closer Look at Shapley Values and LOCO

Collaborative supply chain network design under demand uncertainty: A robust optimization approach

This paper studies a collaborative robust supply chain network design (CRSCND) problem aimed at maximizing economic and social benefits by enabling enterprises to jointly address demand uncertainties. Through strategies including joint inventory replenishment, shared distribution centers (DCs), and pooled transportation resources, the CRSCND problem seeks to optimize plant and DC locations and the allocation of DCs to customers under a collaborative framework. To address this, we develop two robust optimization models incorporating a budget uncertainty set, each model representing a distinct risk-pooling policy. These models are then reformulated into solvable linear programming structures. Results from numerical experiments confirm the cost-reduction benefits of collaboration and robust optimization. Sensitivity analysis reveals that factors like violated probability and high demand volatility minimally impact cost savings enabled by collaboration and robustness. Moreover, each robust model shows distinct suitability depending on specific scenario parameters. Finally, we test three cost-saving allocation mechanisms, finding that only the Shapley value method yields best allocations in cases involving overlapping demand.

Distributed Low-Carbon Economic Dispatching for Multiple Park-level Integrated Energy Systems Based on Improved Shapley Value

Distributed Low-Carbon Economic Dispatching for Multiple Park-level Integrated Energy Systems Based on Improved Shapley Value

Explainable Tuned Machine Learning Models for Assessing the Impact of Corrosion on Bond Strength in Concrete

This study mainly aims to evaluate the bond strength of corroded reinforcements in reinforced concrete members. In this regard, a comprehensive dataset containing a total of 285 specimens was collected from previous experiments. All collected specimens, including normal concrete, were subjected to pull-out tests to ensure consistent results. The features evaluated are associated with both concrete and rebar characteristics, corrosion rate, and duration. Six machine learning (ML) models were used to assess the dataset: Decision Tree, Random Forest, Light Gradient-Boosting Machine, Gradient Boosting, Extreme Gradient Boosting, and Extra Tree. Hyperparameter tuning was conducted using grid search to optimize model performance and enhance predictive accuracy. Additionally, the Shapley Values technique was utilized to interpret the significance of the features on bond strength.The results show that Extreme Gradient Boosting and Extra tree methods outperformed the other models, with score of 0.9 each and RSME of 2.21 and 1.87, respectively. Furthermore, tuned models resulted in more accurate performance than the default models. Evaluating the significance of studied features indicated that the elevated levels of corrosion were associated with a negative impact on bond strength. In addition, the corrosion rate is considered to be the most influential factor affecting the bond strength.

Coordinated Optimization of Integrated Rural Multiple Regional Energy Systems Considering Electricity to Ammonia and Improved Shapley Value Revenue Allocation

Coordinated Optimization of Integrated Rural Multiple Regional Energy Systems Considering Electricity to Ammonia and Improved Shapley Value Revenue Allocation

Degree of Causal Contribution and Degree of Moral Responsibility: Using Shapley Values as a Measure

This paper is about the pros and cons of measuring degree of causal contribution by Shapley Value (a concept from cooperative game theory). The Shapley Value approach defines degree of causal contribution for a causal factor in terms of a kind of average of its marginal contributions across a range of possible groupings of factors. Some existing literature already explores applying Shapley Values to degrees of causal contributions. This paper supplements this literature by evaluating a range of justifications for applying Shapley Values, addressing several potential problems for the Shapley Value approach, and tying this to recent discussions of the tension between degree of sufficiency and degree of necessity originating with Bernstein.

Clinical information extraction for lower-resource languages and domains with few-shot learning using pretrained language models and prompting

Abstract A vast amount of clinical data are still stored in unstructured text. Automatic extraction of medical information from these data poses several challenges: high costs of clinical expertise, restricted computational resources, strict privacy regulations, and limited interpretability of model predictions. Recent domain adaptation and prompting methods using lightweight masked language models showed promising results with minimal training data and allow for application of well-established interpretability methods. We are first to present a systematic evaluation of advanced domain-adaptation and prompting methods in a lower-resource medical domain task, performing multi-class section classification on German doctor’s letters. We evaluate a variety of models, model sizes (further-pre)training and task settings, and conduct extensive class-wise evaluations supported by Shapley values to validate the quality of small-scale training data and to ensure interpretability of model predictions. We show that in few-shot learning scenarios, a lightweight, domain-adapted pretrained language model, prompted with just 20 shots per section class, outperforms a traditional classification model, by increasing accuracy from $48.6\%$ to $79.1\%$ . By using Shapley values for model selection and training data optimization, we could further increase accuracy up to $84.3\%$ . Our analyses reveal that pretraining of masked language models on general-language data is important to support successful domain-transfer to medical language, so that further-pretraining of general-language models on domain-specific documents can outperform models pretrained on domain-specific data only. Our evaluations show that applying prompting based on general-language pretrained masked language models combined with further-pretraining on medical-domain data achieves significant improvements in accuracy beyond traditional models with minimal training data. Further performance improvements and interpretability of results can be achieved, using interpretability methods such as Shapley values. Our findings highlight the feasibility of deploying powerful machine learning methods in clinical settings and can serve as a process-oriented guideline for lower-resource languages and domains such as clinical information extraction projects.

Integrating predictive modeling and causal inference for advancing medical science

Artificial intelligence (AI) is revolutionizing healthcare by providing tools for disease prediction, diagnosis, and patient management. This review focuses on two key AI methodologies in healthcare: predictive modeling and causal inference. Predictive models excel in identifying patterns to forecast outcomes but are limited in explaining the underlying causes. In contrast, causal inference focuses on understanding cause-and-effect relationships, which makes effective medical interventions possible. Although randomized controlled trials (RCTs) are the gold standard for causal inference, they face limitations including cost and ethical concerns. As alternatives, emulated RCTs and advanced machine learning techniques have emerged for estimating causal effects, bridging the gap between prediction and causality. Additionally, Shapley values and Local Interpretable Model-Agnostic Explanations improve the interpretability of complex AI models, making them more actionable in clinical settings. Integrating prediction and causal inference holds great promise for advancing personalized medicine, enhancing patient outcomes, and optimizing healthcare delivery. However, careful application of AI tools is crucial to avoid misinterpretation and maximize their potential.

Dynamic Operational Planning in Warfare: A Stochastic Game Approach to Military Campaigns

ABSTRACTWe study a two‐player discounted zero‐sum stochastic game model for dynamic operational planning in military campaigns. At each stage, the players manage multiple commanders who order military actions on objectives that have an open line of control. When a battle over the control of an objective occurs, its stochastic outcome depends on the actions and the enabling support provided by the control of other objectives. Each player aims to maximize the cumulative number of objectives they control, weighted by their criticality. To solve this large‐scale stochastic game, we derive properties of its Markov perfect equilibria by leveraging the logistics and military operational command and control structure. We show the consequential isotonicity of the optimal value function with respect to the partially ordered state space, which in turn leads to a significant reduction of the state and action spaces. We also accelerate Shapley's value iteration algorithm by eliminating dominated actions and investigating pure equilibria of the matrix game solved at each iteration. We demonstrate the computational value of our equilibrium results on a case study that reflects representative operational‐level military campaigns with geopolitical implications. Our analysis reveals a complex interplay between the game's parameters and dynamics in equilibrium, resulting in new military insights for campaign analysts.

Gately values of cooperative games

AbstractWe investigate Gately’s solution concept for cooperative games with transferable utilities. Gately’s conception introduced a bargaining solution that minimises the maximal quantified “propensity to disrupt” the negotiation process of the players over the allocation of the generated collective payoffs. We show that Gately’s solution concept is well-defined for a broad class of games and that it can be interpreted as a compromise solution. We also consider a generalisation based on a parameter-based quantification of the propensity to disrupt. We provide an axiomatic characterisation of the original Gately value as well as these generalised Gately values. Furthermore, we investigate the relationship of these Gately values with the Core and the Nucleolus and show that Gately’s solution is in the Core for all regular 3-player games, but is fundamentally different from the Nucleolus. We identify exact conditions under which these Gately values are Core imputations for arbitrary regular cooperative games. Finally, we investigate the relationship of the Gately value with the Shapley value.