Musashino University

SHapley Additive Explanations (SHAP) are widely used to interpret machine learning models in agriculture and environmental decision-making, but SHAP inherits model misspecification, confounding, and distribution shift, risking unstable and policy-misleading importance rankings. This opinion advances a concrete, model-agnostic workflow that goes beyond SHAP: pair unsupervised structure checks (e.g., feature agglomeration, highly variable feature selection) with nonlinear, nonparametric association metrics and pre-registered sensitivity/stability analyses, then audit attributions with domain-informed negative controls and decision impact tests. The novelty lies in reframing SHAP from a standalone explainer to one component of a robustness protocol explicitly designed to reduce label-driven and proxy bias. This integrated approach yields more reliable variable importance, tighter uncertainty communication, and clearer links between attribution changes and real policy choices.

Marine Pollution Bulletin increasingly applies machine learning and explainable AI to pollutant and shellfish poisoning risk, exemplified by PCA-based source apportionment and SHAP-based feature attribution. However, linear PCA may misrepresent structure in inherently nonlinear environmental data, and existing studies often treat model-derived feature importances as evidence of true associations without assessing consistency or dose-response relationships. This paper clarifies that supervised models possess two distinct accuracies: prediction and feature importance, and only prediction can be validated against ground truth. Using a Basque coastal dataset (8195 instances, 14 features) with chlorophyll-a as a proxy for paralytic shellfish poisoning risk, we introduce a leave-top1-out procedure to test ranking stability. Random Forest and XGBoost with and without SHAP show pronounced instability, indicating biased, model-dependent importances. In contrast, unsupervised and non-target-prediction methods yield perfectly stable rankings while matching or exceeding supervised performance, supporting routine stability, consistency, dose-response, and linearity checks in environmental ML studies.

In this article, we investigate the blow-up behavior of solutions to the one-dimensional damped nonlinear wave equation, namely ∂_t^2 u - ∂_x^2 u + 1 + t∂_t u = |∂_t u|^p (p > 1). Under the assumption of sufficiently large and smooth initial data, we establish that the blow-up curve is continuously differentiable (𝒞^1). A key step in our analysis involves the characterization of the blow-up profile of the solution. The proof relies on transforming the equation into a first-order system and adapting the techniques of Sasaki in which have elegantly extended the method of Caffarelli and Friedman to nonlinear wave equations with time derivative nonlinearity, but without the scale-invariant term (μ=0).

AIM:Hikikomori is a state of prolonged social isolation in which individuals remain at home and disengage from work, school, and other activities. This preliminary study investigated the correspondence between self-reported scores on the Hikikomori Functional Assessment Scale (HFAS) and independently collected interview-based ratings. Consistency was evaluated at two retrospectively defined time points: the participant-designated peak period of hikikomori (Peak) and the time of assessment (Current). METHODS:A total of 16 participants (6 men and 10 women; mean age = 40.2 years) completed the HFAS and a semi-structured interview for both time points. Interviewers were masked to questionnaire responses. Data were analysed using Wilcoxon signed-rank tests and Kendall's τ correlations with Bayes factors. RESULTS:Social negative reinforcement showed moderate to strong correspondence between methods at both time points (Peak: τ = 0.50, BF10 = 8.9; Current: τ = 0.50, BF10 = 8.6). Intrapersonal negative reinforcement indicated weak to moderate correlations and evidence at Peak (τ = 0.40, BF10 = 3.0) but not at Current (τ = 0.07, BF10 = 0.5). Intrapersonal positive reinforcement exhibited only anecdotal evidence of association at either time point (Peak: τ = 0.27, BF10 = 1.1; Current: τ = 0.27, BF10 = 1.0). CONCLUSIONS:These findings provide preliminary support for the HFAS as a practical tool for identifying socially avoidant contingencies relevant to early intervention, underscoring the need for further research with larger, prospective samples.

Background: Artificial intelligence (AI) is becoming important in oncology, supporting risk prediction, treatment planning, and biomarker discovery. However, current evaluation practices often assume that high predictive accuracy implies reliable interpretation-a misconception that may undermine reproducibility and clinical decision-making. This study aims to reassess interpretability by introducing feature ranking order consistency as a stability-focused metric to evaluate how model explanations respond to minimal input perturbations. Methods: Using The Cancer Genome Atlas (TCGA) breast cancer multi-omics dataset, we compared supervised models-Linear Regression, Least Absolute Shrinkage and Selection Operator (LASSO), Random Forest, and Extreme Gradient Boosting (XGBoost)-with unsupervised and statistical methods, including Principal Component Analysis (PCA), Highly Variable Gene Selection, and Spearman's rank correlation. Each method produced a Top 20 feature ranking, and stability was assessed by testing whether rankings remained consistent after removing the top-ranked feature. Predictive performance was evaluated using a Random Forest classifier with stratified 10-fold cross-validation. Results: Supervised models exhibited unstable feature importance rankings even under minimal perturbations (<0.1% feature removal), suggesting that high predictive accuracy may obscure fragile or misleading explanations. In contrast, Highly Variable Gene Selection and Spearman's correlation consistently produced stable, biologically coherent feature sets and maintained competitive predictive performance. Conclusions: Interpretive instability is a major limitation of many machine learning models in oncology. Incorporating stability-based criteria-such as feature ranking consistency-into evaluation frameworks is essential for ensuring reproducible, trustworthy, and clinically actionable AI. As AI adoption accelerates, prioritizing interpretability alongside accuracy is critical for responsible deployment in precision oncology.