武藤 佳恭

This paper outlines key machine learning principles, focusing on the use of XGBoost and SHAP values to assist researchers in avoiding analytical pitfalls. XGBoost builds models by incrementally adding decision trees, each addressing the errors of the previous one, which can result in inflated feature importance scores due to the method's emphasis on misclassified examples. While SHAP values provide a theoretically robust way to interpret predictions, their dependence on model structure and feature interactions can introduce biases. The lack of ground truth values complicates model evaluation, as biased feature importance can obscure real relationships with target variables. Ground truth values, representing the actual labels used in model training and validation, are crucial for improving predictive accuracy, serving as benchmarks for comparing model outcomes to true results. However, they do not ensure real associations between features and targets. Instead, they help gauge the model's effectiveness in achieving high accuracy. This paper underscores the necessity for researchers to recognize biases in feature importance and model evaluation, advocating for the use of rigorous statistical methods to enhance the reliability of analyses in machine learning research.

Gomez-Flores et al. proposed a Long Short-Term Memory Neural Network (LSTM-NN) for predicting the flotation behavior of battery active materials using various physicochemical and hydrodynamic variables. While they achieved high prediction accuracy, validated through Mean Relative Error (MRE) and Mean Squared Error (MSE) metrics, concerns arise regarding the integrity of feature importance assessments derived from SAGE and SHAP methodologies. Specifically, the reliance on these model-specific techniques can introduce biases, obscuring the true relationships between features. Additionally, while Spearman's correlation elucidated significant relationships among variables, the absence of discussion on p-values left gaps in interpretation. This study emphasizes the need for cautious interpretation of feature importance metrics and the elimination of less significant variables, aiming to enhance model robustness and improve actionable insights in machine learning contexts.

This study analyzes trends in felony sentence disparity based on gender and race from 2010 to 2024. It utilizes a generative AI to create Python code for data visualization and employs three statistical methods (ANOVA, Chi-Square, Fisher's Exact) to assess p-values. The p-value signifies the probability of random chance causing the observed association. A significance level of 0.05 is used as a benchmark. The evidence-based analysis reveals a concerning trend: increasing disparities in sentences across genders and races. The findings highlight the need for further research and policy changes to address these disparities in the criminal justice system. The paper offers a novel visualization approach to depict these trends, aiding comprehension of the issue.

Cultural norms and traditional behaviors have significantly influenced the outcomes of the COVID-19 pandemic. Asian countries initially outperformed their Western counterparts due to their cultural practices. However, policy shifts have led to a decline in these countries' performance. The objective of this study is to scrutinize the performance of different countries in managing the COVID-19 pandemic, with a focus on the influence of cultural norms and traditional behaviors, and to propose a tool that can inform and enhance current urban management policies. This study employs a time-series policy outcome analysis tool that operates on a single metric: the daily cumulative mortality of the population. By implementing a test-isolation strategy to manage quarantine periods, this tool aims to significantly influence the pandemic's outcome. The tool's efficacy is showcased through a case study involving four countries. New insights are validated and visualized via generated graphs, demonstrating the potential of this tool in the realm of tourism and urban management. This proposed tool holds promise for informing and enhancing current urban management policies, thereby mitigating unnecessary tourism-related fatalities in future pandemics. It underscores the importance of having the right information at the right time to make informed decisions in response to a pandemic.

This paper critically examines the analysis conducted by Maußner et al. on AI analysis, particularly their interpretation of feature importances derived from various machine learning models using SHAP (SHapley Additive exPlanations). Although SHAP aids in interpretability, it is subject to model-specific biases that can misrepresent relationships between variables. The paper emphasizes the lack of ground truth values in feature importance assessments and calls for careful consideration of statistical methodologies, including robust nonparametric approaches. By advocating for the use of Spearman's correlation with p-values and Kendall's tau with p-values, this work aims to strengthen the integrity of findings in machine learning studies, ensuring that conclusions drawn are reliable and actionable.

This study investigates the trends of Invasive Pneumococcal Disease (IPD) in the US from 1998 to 2021, segmented by age groups, using data from the CDC. The trends were visualized using generative AI, revealing a correlation between the IPD trends in the US and the COVID-19 pandemic. A significant reduction in the number of cases for serotype 14 was observed in the “< 2” age group, which decreased from 246 cases in 2000–58 cases in 2001. This drop could be a contributing factor to the overall decline in cases during this period. Similarly, in the “65 + ” age group, a significant reduction in the number of cases for serotype 3 was noticed, which fell from 197 cases in 2019–64 cases in 2021. These findings underscore the importance of continuous monitoring and intervention strategies in managing IPD.

This literature review examines the interplay between myofascial release, mental health, and sleep quality, utilizing peer-reviewed studies from the National Library of Medicine (2021–2024). Analyzing ten selected articles, findings indicate that myofascial release significantly enhances mental health by alleviating depression and anxiety associated with various physical conditions, such as premenstrual syndrome, chronic pain syndromes, postpartum dysfunctions, and fibromyalgia. These improvements in mental well-being are closely linked to enhanced sleep quality and overall quality of life. For instance, myofascial techniques combined with progressive relaxation improved sleep and reduced pain in premenstrual syndrome, while similar approaches in chronic low back pain and fibromyalgia patients led to better sleep and reduced psychological distress. The review underscores myofascial release as a promising complementary therapy for promoting mental health and sleep, though it calls for more rigorous randomized-controlled trials to further validate these effects and elucidate underlying mechanisms.

This paper explores a novel approach using generative AI to enhance drug marketing strategies in the US pharmaceutical sector. By leveraging an official dataset sourced from the US government, the AI generates Python code to analyze the time interval between FDA approval dates and market release dates. The analysis identifies 370 manufacturers who achieved "zero-day" marketing-referring to drugs marketed immediately upon FDA approval-and 174 manufacturers who marketed their products within less than seven days of approval. Notably, 947 drug products were found to have been marketed prior to FDA approval, raising significant regulatory and ethical concerns that necessitate further discussion. The findings indicate that 174 drug manufacturers have the potential to optimize their marketing strategies to achieve zero-day timelines, prompting an examination of the feasibility of such acceleration within the current regulatory framework and its implications for industry practices. Additionally, this paper discusses the broader impact of AI-driven strategies in the pharmaceutical sector, highlighting their potential to not only enhance marketing speed but also improve aspects such as compliance and decision-making efficiency. Furthermore, a tutorial on implementing generative AI is provided, detailing how it can be utilized to achieve marketing objectives through interactive conversations with the AI. This practical application demonstrates the technology's capabilities using real dataset analysis and reveals significant findings that could inform future strategies within the industry. The research objectives and their broader implications underscore the need for ongoing dialogue about the ethical and regulatory dimensions of AI in pharmaceutical marketing.

Background: Zheng et al. (2025) provided a comprehensive review of advancements in electronic noses used for detecting alcoholic beverages. Their work highlights the critical role of Principal Component Analysis (PCA) in feature reduction, which enhances the accuracy of various analytical methods such as linear discriminant analysis (LDA), random forest (RF), convolutional neural networks (CNN), and back propagation neural networks (BPNN). While PCA is a widely used technique, its application in electronic nose technologies necessitates a closer examination of its limitations. Scope and approach: This paper critically evaluates the limitations of PCA when applied to nonlinear and nonparametric data, emphasizing the potential for distorted conclusions that can arise from its use. Through an extensive literature review, the paper discusses the implications of PCA within electronic nose applications. Key areas of focus include the importance of assessing data distribution, understanding statistical relationships, and validating significance using p-values. Additionally, the paper advocates for the adoption of nonparametric statistical methods, such as Spearman's correlation and Kendall's tau, to enhance the reliability of the analyses conducted. Key findings and conclusion: The review reveals that the linear assumptions underlying PCA may misrepresent variance in nonlinear datasets, leading to misleading projections that obscure structural information. PCA's focus on global patterns can also overlook significant local variations, potentially causing overlaps among distinct classes within high-dimensional data. These limitations necessitate caution when utilizing PCA in electronic nose technologies. Therefore, to ensure valid and reliable results in this rapidly advancing field, it is essential to adopt robust statistical methods and conduct thorough preliminary analyses that account for the specific characteristics of the data. Mitigating the risks of distorted conclusions will improve the accuracy and credibility of findings in this area of research.

This study aimed to evaluate the prevalence of dental implants and the factors influencing their survival rates, including systemic disorders, medication use, lifestyle habits, and implant design. A literature review revealed that implants with laser-microtextured grooves exhibited lower peri‑implantitis incidence and higher survival rates. Early failure often correlated with smoking, male gender, and younger age, while adjacent teeth faced an increased risk of loss. Personality traits were found to affect implant success in older patients, alongside concerns regarding the durability of titanium implants. The findings stress the necessity of comprehensive patient evaluations and enhanced diagnostic skills for improving dental implant outcomes.

Background: Bhat et al. (2025) highlight the significant role of artificial intelligence (AI) and machine learning (ML) in food authentication through advanced algorithms that analyze large datasets for patterns associated with food fraud. Objective: This paper aims to critically assess the approach of Bhat et al., with a specific focus on model-based feature importance and the biases related to traditional machine learning methods. Methods: The paper distinguishes between machine learning target predictions and feature importances, advocating for the rigorous application of robust statistical techniques, including Spearman's correlation and p-values, to accurately reveal genuine associations among variables. Results: The analysis emphasizes the necessity for researchers to comprehend the foundational principles of AI and ML to avoid misapplication of these technologies. Conclusion: The paper recommends integrating both nonparametric and nonlinear methods to effectively reduce bias and improve the reliability of feature importance assessments in food authentication.

This study tackles creating Python code for beginners with generative AI and analyzing trends in mental distress among Alzheimer's patients in the US (2015–2022 CDC data). It guides beginners through using AI to generate code for visualizing these trends by age and sex. The findings reveal females, particularly those 50–64 years old, experience the highest rates of mental distress. This emphasizes the importance of considering age and sex when developing care and interventions for mental distress in Alzheimer's patients.

Han and Lin's phase field model for lithium-ion batteries utilizes LASSO regression to analyze battery performance during galvanostatic cycling, aiming to simplify the relationship between parameters and Coulombic efficiency. Despite demonstrating accuracy, this paper critiques the reliance on LASSO for feature selection, highlighting its potential inadequacy in capturing nonlinear interactions within chemical properties. Traditional performance metrics, such as MAE, RMSE, and R², provide limited insights regarding individual feature contributions and the nuanced relationships present in the data. This paper advocates for adopting nonparametric statistical methods, such as Spearman's correlation and Kendall's tau, which can better elucidate complex variable associations and validate feature importance. Incorporating these methods will enhance the robustness of findings, promoting a clearer understanding of battery performance dynamics.

Skorobogatov et al. developed supervised machine learning models to predict diagnoses and illness states in schizophrenia and bipolar disorder. However, their reliance on bootstrap forests and generalized regressions introduces significant biases in feature importance assessments. This paper highlights the critical distinction between feature importances generated by machine learning and actual associations, which are often model-specific and context-dependent. We underscore the limitations of biased feature importances and advocate for the use of robust statistical methods, such as Chi-squared tests and Spearman's correlation, to reveal true associations. Reassessing findings with these methods will enable more accurate interpretations and reinforce the importance of understanding the limitations inherent in machine learning methodologies.

Burnout among rheumatologists is globally prevalent, driven by low personal accomplishment, younger age, dissatisfaction with the specialty, low income, long hours, emotional exhaustion, and depersonalization. Mitigation strategies include addressing modifiable risk factors, implementing organizational measures, investing in well-being, assessing individual grit, and managing workload with virtual care platforms.

Mohan et al. developed a feed-forward neural network (FFNN) model to predict Kamlet-Taft parameters using quantum chemically derived features, achieving notable predictive accuracy. However, this study raises concerns about conflating prediction accuracy with feature importance accuracy, as high R2 and low RMSE do not guarantee valid feature importance assessments. The reliance on SHAP (SHapley Additive exPlanations) for feature evaluation is problematic due to model-specific biases that could misrepresent true associations. A broader understanding of data distribution, statistical relationships, and significance testing through p-values is essential to rectify this. This paper advocates for employing robust statistical methods, like Spearman's correlation, to effectively assess genuine associations and mitigate biases in feature importance analysis.

The veterinary profession faces a critical challenge: burnout. Long hours, emotional strain, financial pressures, and difficult client interactions contribute to stress and drive veterinary professionals from the field. This harms not only their well-being but also patient care and workplace morale. Research highlights the concerning mental state of veterinarians, with studies finding high rates of burnout, compassion fatigue, and burden transfer (stress from client challenges). Early-career and female veterinarians are most vulnerable. Several studies explore interventions to improve veterinary well-being. Educational programs targeting communication and acceptance of difficult clients show promise in reducing stress and burnout. Additionally, a web-based acceptance and commitment therapy (ACT) program demonstrates effectiveness in reducing occupational distress. It is crucial to address the veterinary burnout crisis. By creating supportive work environments, prioritizing mental health, and implementing interventions like ACT training, we can retain veterinary professionals and ensure a thriving profession that provides optimal care for animals and their caregivers.

Nutraceuticals, with their potential health benefits, are increasingly being used to manage a variety of health conditions. The global market for nutraceuticals, valued at USD 540 billion in 2022, is projected to reach USD 1025 billion by 2030. This paper delves into the beneficial impacts of emerging nutraceuticals on a spectrum of medical disorders, drawing from credible sources from the National Library of Medicine. We have scrutinized studies on the application of nutraceuticals in treating conditions like sleep disorders, migraines, oxidative stress, mental health issues, pain disorders, obesity, gastrointestinal disorders, and even COVID-19. Our analysis indicates that nutraceuticals hold promise for addressing various health issues. However, this paper also sheds light on the health risks associated with nutraceuticals. Despite their widespread use, the safety and efficacy of nutraceuticals are still uncertain due to the lack of stringent regulations, unlike pharmaceutical drugs. This raises concerns about potential health risks and misleading claims. Research indicates that some supplements can cause adverse effects and interact with medications. Therefore, to ensure safe usage, it is imperative to implement stricter regulations, enhance reporting systems, and boost consumer awareness.

The study analyzes a CDC dataset on US adult obesity and physical activity from 2011 to 2022. Despite rising obesity rates and insufficient fruit consumption, physical activity levels are increasing and overweight rates are slightly declining. The role of ultra-processed food intake, price sensitivity, early eating habits, and stress in obesity is highlighted. The findings suggest a complex obesity epidemic, indicating the need for multifaceted solutions such as regulating ultra-processed foods, improving access to healthy foods, and promoting healthy eating habits from childhood.

Annually, a tragic toll of 1.3 million lives is lost on roads across the globe, with tens of millions more suffering injuries or disabilities. The necessity for precise detection of abnormal driving behavior is paramount in reducing traffic accidents. This paper aims to bridge the gap between normal and abnormal driving patterns, offering near-flawless detection capabilities. This paper presents a novel AI tachograph prototype, the first of its kind, that can classify driving behavior into normal and abnormal in real time with an impressive accuracy of 99.99 %. This high level of accuracy is achieved by using a bias-reduction method. The bias-reduction method focuses on minimizing biases in the dataset, such as surrounding situations, location, driver information, and car types. This approach significantly enhances the prediction accuracy of existing machine learning algorithms. The dataset used for this research is quite extensive, consisting of anomaly data collected from 10,181 commercial vehicles and 12,530 drivers in just 0.1 s. This rich dataset is crucial for building a reliable model. The effectiveness of the proposed method was validated using 10-fold cross validation on 480 k to 540 k instances with 36 determinants. The results clearly demonstrated that reducing bias leads to higher prediction accuracy. The paper also plans to compare the prediction accuracy of balanced and imbalanced datasets. The findings from this research have broader implications as the proposed method can be applied generally to machine learning to improve prediction accuracy.

This review examines how various medications can trigger inflammation throughout the body. It explores causes, ranging from common pain relievers like NSAIDs to chemotherapy drugs. The review also highlights potential treatments, including established medications and promising new therapies. Physicians and patients can work together to reduce this risk by understanding these causes and implementing preventive measures, such as monitoring for side effects and using alternative medications when possible. Drug-induced inflammation can be categorized into four types based on the immune response involved. Symptoms vary by type and affected organ. Common symptoms include fever, malaise, joint pain, rash, and swelling. Diagnosis involves blood tests, imaging, and biopsies. Treatment primarily involves discontinuing the suspected drug and providing supportive care. The development of new drugs and therapies has made diagnosis challenging. However, recent advances in biomarkers and genetic risk assessment techniques are improving diagnosis and risk assessment of drug-induced liver injury. Preventive measures for drug-induced inflammation include monitoring for side effects, using alternative medications, developing new drug delivery methods, exploring new anti-inflammatory drugs, being aware of rare side effects, and understanding the underlying mechanisms.

This paper demonstrates that the conclusions drawn from datasets on global temperature anomaly and atmospheric CO2 from NOAA can vary depending on the range of investigated periods. By examining the data from both macroscopic and microscopic perspectives, the study reveals that different levels of analysis can produce different outcomes from the same datasets based on statistics.

Instructors are always interested in methods to activate learner incentives and motivation to increase learning effectiveness. This paper introduces the Python Package Index (PyPI) as a powerful tool to maximize learner incentives on software and presents an example of its application in agriculture and science. The more useful the PyPI application is, the more it will be downloaded worldwide, providing an external review for the learner, and strengthening their incentive. However, many existing tutorials on PyPI, including the official site, are not updated on the twine library for uploading files to the PyPI site. This paper presents an updated tutorial on using PyPI for counting disaggregated objects such as bugs and pests, and for software reproducibility validation via Code Ocean. Additionally, generative AI is introduced as an indispensable assistant for tasks such as understanding technical terms and providing solutions for encountered problems.

This tutorial, rooted in the context of livestock research, is designed to assist novice or non-programmers in visualizing trends in livestock exports between the US and Japan using Python and generative AI systems such as Microsoft's Copilot and Google's Gemini. The analysis of these trends plays a pivotal role in optimizing livestock production. The tutorial offers a thorough guide on preparing data using reliable federal datasets, generating Python code, and tackling potential issues such as overlapping data points. It effectively simplifies complex tasks into manageable steps and includes Python code in the appendices for easy reference. By enabling researchers to extract insights and make predictions from livestock data, this tutorial addresses a significant void in the existing literature. This innovative approach has the potential to transform the way researchers engage with and interpret livestock data, thereby making a substantial contribution to the field.