Feature Importance Analysis Research Articles

Experimentation of AI Models Towards the Prediction of Medium-Risk Emergency Department Cases Disposition Outcome.

The overwhelming volume of patients in emergency departments (EDs) is a significant problem that hinders the delivery of high quality healthcare. Despite their great value, triage protocols are challenging to put into practice. This paper examines the utility of prediction models as a tool for clinical decision support, with a focus on medium-severity patients as defined by the ESI algorithm. 689 cases of medium-risk patients were gathered from the AHEPA hospital, evaluated, and their data fed three classifiers: XGBoost (XGB), Random Forest (RF) and Logistic Regression (LR), with the prediction goal being the outcome of their visit, i.e. admission or discharge. Essential features for the prediction task were determined using feature importance and distribution analysis. Despite having many missing values or high sparsity datasets, several symptoms and metrics are recommended as crucial for outcome prediction. When fed the patients' vital signs, XGB achieved an accuracy score of 91.30%. Several chief complaints were also proven beneficial. Prediction models can, in general, not only lessen the drawbacks of triage implementation, but also enhance its delivery.

Exploring Prediabetes Pathways Using Explainable AI on Data from Electronic Medical Records.

This study leverages data from a Canadian database of primary care Electronic Medical Records to develop machine learning models predicting type 2 diabetes mellitus (T2D), prediabetes, or normoglycemia. These models are used as a basis for extracting counterfactual explanations and derive personalized changes in biomarkers to prevent T2D onset, particularly in the still reversible prediabetic state. The models achieve satisfactory performance. Furthermore, feature importance analysis underscores the significance of fasting blood sugar and glycated hemoglobin, while counterfactuals explanations emphasize the centrality of keeping body mass index and cholesterol indicators within or close to the clinically desirable ranges. This research highlights the potential of machine learning and counterfactual explanations in guiding preventive interventions that may help slow down the progression from prediabetes to T2D on an individual basis, eventually fostering a recovery from prediabetes to a normoglycemic state.

Enhancing Depression Detection: A Stacked Ensemble Model with Feature Selection and RF Feature Importance Analysis Using NHANES Data

Around the world, 5% of adults suffer from depression, which is often inadequately treated. Depression is caused by a complex relationship of cultural, psychological, and physical factors. This growing issue has become a significant public health problem globally. Medical datasets often contain redundant characteristics, missing information, and high dimensionality. By using an iterative floating elimination feature selection algorithm and considering various factors, we can reduce the feature set and achieve optimized outcomes. The research utilizes the 36-Item Short Form Survey (SF-36) from the NHANES 2015–16 dataset, which categorizes data into seven groups relevant to quality of life and depression. This dataset presents a challenge due to its imbalance, with only 8.08% of individuals diagnosed with depression. The Depression Ensemble Stacking Generalization Model (DESGM) employs stratified k-fold cross-validation and oversampling for training data. DESGM enhances the classification performance of both base learners (linear support vector machine, perceptron, artificial neural network, linear discriminant analysis, and K-nearest neighbor) and meta-learners (logistic regression). The model achieved an F1 score of 0.9904 and an accuracy of 98.17%, with no instances of depression misdiagnosed.

Templated Nucleation of Clotrimazole and Ketoprofen on Polymer Substrates.

The use of different template surfaces in crystallization experiments can directly influence the nucleation kinetics, crystal growth, and morphology of active pharmaceutical ingredients (APIs). Consequently, templated nucleation is an attractive approach to enhance crystal nucleation kinetics and preferentially nucleate desired crystal polymorphs for solid-form drug molecules, particularly large and flexible molecules that are difficult to crystallize. Herein, we investigate the effect of polymer templates on the crystal nucleation of clotrimazole and ketoprofen with both experiments and computational methods. Crystallization was carried out in toluene solvent for both APIs with a template library consisting of 12 different polymers. In complement to the experimental studies, we developed a computational workflow based on molecular dynamics (MD) and derived descriptors from the simulations to score and rank API-polymer interactions. The descriptors were used to measure the energy of interaction (EOI), hydrogen bonding, and rugosity (surface roughness) similarity between the APIs and polymer templates. We used a variety of machine learning models (14 in total) along with these descriptors to predict the crystallization outcome of the polymer templates. We found that simply rank-ordering the polymers by their API-polymer interaction energy descriptors yielded 92% accuracy in predicting the experimental outcome for clotrimazole and ketoprofen. The most accurate machine learning model for both APIs was found to be a random forest model. Using these models, we were able to predict the crystallization outcomes for all polymers. Additionally, we have performed a feature importance analysis using the trained models and found that the most predictive features are the energy descriptors. These results demonstrate that API-polymer interaction energies are correlated with heterogeneous crystallization outcomes.

Explainable machine learning-driven predictive performance and process parameter optimization for caproic acid production

In this study, four machine learning (ML) prediction models were developed to predict and optimize the production performance of caproic acid based on substrates, products, and process parameters. The XGBoost outperformed others, with a high R2 of 0.998 on the training set and 0.885 on the test set. Feature importance analysis revealed hydraulic retention time (HRT) and butyric acid concentration are decisive. The SHAP method offered profound insights into the interplay and cumulative effects of substrate composition, identified the synergistic effects between butyric acid and lactic acid, and emphasized adding glucose can benefit caproic with lactic acid co-fermentation. By integrating the Adaptive Variation Particle Swarm Optimization (AVPSO) algorithm, the optimal process conditions to achieve a maximum caproic acid production of 8.64 g/L was obtained. This study not only advances caproic acid production but contributes a versatile ML-driven strategy applicable to bioprocess optimizations, potentially transformative for sustainable and economically viable bioproduction.

Predicting High‐Performance Thermoelectric Materials With StarryData2

AbstractIn recent years, machine learning (ML) has emerged as a potential tool in the exploration of thermoelectric (TE) materials. This study exploits the StarryData2 public database to construct an ML model for predicting the figure‐of‐merit ZT of TE materials. The original dataset from StarryData2 (372,480 datapoints) underwent systematic cleaning, resulting in a refined dataset of 18,126 instances with 2,761 unique compounds. The cleaned data is employed to train an XGBoost regressor model, utilizing chemical formulas of TE compounds as features to predict ZT at given temperatures. The XGBoost regressor exhibited high prediction accuracy, achieving the coefficient of determination (R2) scores of 0.815 and mean absolute error (MAE) of 0.103 for the test set, further evaluated through cross‐validation across 5 folds. The learning curve analysis demonstrated improved model performance with increased training data. Furthermore, the contributions of different chemical descriptors to ZT are analyzed based on feature importance analysis. Beyond conventional TE families in the training set, the trained model is applied to predict ZT for promising unexplored TE materials and estimate optimal doping concentrations. This comprehensive study shows the impact of ML on TE material research, offering valuable insights and accelerating the discovery of materials with enhanced TE properties.

Common and differential variables of anxiety and depression in adolescence: a nation-wide smartphone-based survey

BackgroundMental health in adolescence is critical in its own right and a predictor of later symptoms of anxiety and depression. To address these mental health challenges, it is crucial to understand the variables linked to anxiety and depression in adolescence.MethodsHere, we analyzed data of 278 adolescents that were collected in a nation-wide survey provided via a smartphone-based application during the COVID-19 pandemic. We used an elastic net regression machine-learning approach to classify individuals with clinically relevant self-reported symptoms of depression or anxiety. We then identified the most important variables with a combination of permutation feature importance calculation and sequential logistic regressions.Results40.30% of participants reported clinically relevant anxiety symptoms, and 37.69% reported depressive symptoms. Both machine-learning models performed well in classifying participants with depressive (AUROC = 0.77) or anxiety (AUROC = 0.83) symptoms and were significantly better than the no-information rate. Feature importance analyses revealed that anxiety and depression in adolescence are commonly related to sleep disturbances (anxiety OR = 2.12, depression OR = 1.80). Differentiating between symptoms, self-reported depression increased with decreasing life satisfaction (OR = 0.43), whereas self-reported anxiety was related to worries about the health of family and friends (OR = 1.98) as well as impulsivity (OR = 2.01).ConclusionOur results show that app-based self-reports provide information that can classify symptoms of anxiety and depression in adolescence and thus offer new insights into symptom patterns related to adolescent mental health issues. These findings underscore the potentials of health apps in reaching large cohorts of adolescence and optimize diagnostic and treatment.

Determinants of Food Security in Rural Households: An Analysis of Dietary Diversity, Land Ownership, and Socioeconomic Factors

Background: Food security is a critical issue in rural areas, influenced by various socioeconomic factors, dietary practices, and resource ownership. Understanding the determinants of food security can help in designing effective interventions to improve the well-being of rural households. Purpose: This study aims to identify and analyze the key determinants of food security among rural households, focusing on dietary diversity, land ownership, and other socioeconomic factors. Methods: The dataset used in this study includes variables such as gender, marital status, education, land ownership, food crop cultivation, meal frequency, age, food source, Food Security Index (FSI), and Household Dietary Diversity Score (HDDS). Multiple linear regression analysis was conducted to explore the relationships between these variables and the FSI. Additionally, a Random Forest Regressor model was employed to predict the FSI and to determine the importance of each feature. Results: The multiple linear regression analysis revealed that HDDS has a significant positive relationship with the FSI (p<0.001), indicating that higher dietary diversity is associated with higher food security levels. Other variables, such as education and food source, showed weaker correlations with food security. The Random Forest Regressor model achieved an R-squared value of 0.495, with feature importance analysis indicating that HDDS, food source, and age are the most influential factors in determining food security. The Mean Squared Error (MSE) of the model was 10.743. Conclusion: This research highlights the crucial role of dietary diversity and socioeconomic conditions in shaping food security outcomes in rural areas. The findings provide valuable insights for policymakers and stakeholders aiming to enhance food security and dietary quality in rural communities. Further studies are recommended to explore the complex interactions between these variables and to develop targeted interventions to improve food security.

Machine learning models for predicting water quality index: optimization and performance analysis for El Moghra, Egypt

ABSTRACT Assessing groundwater quality is vital for irrigation, but financial constraints in developing countries often result in infrequent sampling. This study comprehensively analyzes the groundwater quality of the El Moghra aquifer in Egypt's arid Western Desert, for its suitability for irrigation uses. Detailed hydrochemical analysis and advanced machine learning (ML) techniques, including geographic information systems, were employed to enhance spatial analysis and predictive accuracy. Various ML models, such as random forest, adaptive boosting, and extreme gradient boosting (XGBoost), were optimized using Bayesian optimization to predict the irrigation water quality index (IWQI) accurately. The evaluation incorporated visual and quantitative methods, alongside ranking analysis, to validate model effectiveness. Shapley Additive exPlanations feature importance analysis and a graphical user interface (GUI) model were developed based on the best predictive model. The results indicated that the groundwater quality is generally suitable for irrigation, with XGBoost showing the best performance, achieving a root mean square error of 5.602 and a determination coefficient (R²) of 0.872. Sodium concentration was identified as the most significant factor affecting the IWQI. The GUI facilitates easy prediction of IWQI, aiding agricultural water management and resource allocation within the region.

Predicting tissue distribution and tumor delivery of nanoparticles in mice using machine learning models

Nanoparticles (NPs) can be designed for targeted delivery in cancer nanomedicine, but the challenge is a low delivery efficiency (DE) to the tumor site. Understanding the impact of NPs' physicochemical properties on target tissue distribution and tumor DE can help improve the design of nanomedicines. Multiple machine learning and artificial intelligence models, including linear regression, support vector machine, random forest, gradient boosting, and deep neural networks (DNN), were trained and validated to predict tissue distribution and tumor delivery based on NPs' physicochemical properties and tumor therapeutic strategies with the dataset from Nano-Tumor Database. Compared to other machine learning models, the DNN model had superior predictions of DE to tumors and major tissues. The determination coefficients (R2) for the test datasets were 0.41, 0.42, 0.45, 0.79, 0.87, and 0.83 for DE in tumor, heart, liver, spleen, lung, and kidney, respectively. All the R2 and root mean squared error (RMSE) results of the test datasets were similar to the 5-fold cross validation results. Feature importance analysis showed that the core material of NPs played an important role in output predictions among all physicochemical properties. Furthermore, multiple NP formulations with greater DE to the tumor were determined by the DNN model. To facilitate model applications, the final model was converted to a web dashboard. This model could serve as a high-throughput pre-screening tool to support the design of new and efficient nanomedicines with greater tumor DE and serve as an alternative tool to reduce, refine, and partially replace animal experimentation in cancer nanomedicine research.

Exploring the potential of machine learning to understand the occurrence and health risks of haloacetic acids in a drinking water distribution system

Determining the occurrence of disinfection byproducts (DBPs) in drinking water distribution system (DWDS) remains challenging. Predicting DBPs using readily available water quality parameters can help to understand DBPs associated risks and capture the complex interrelationships between water quality and DBP occurrence. In this study, we collected drinking water samples from a distribution network throughout a year and measured the related water quality parameters (WQPs) and haloacetic acids (HAAs). 12 machine learning (ML) algorithms were evaluated. Random Forest (RF) achieved the best performance (i.e., R2 of 0.78 and RMSE of 7.74) for predicting HAAs concentration. Instead of using cytotoxicity or genotoxicity separately as the surrogate for evaluating toxicity associated with HAAs, we created a health risk index (HRI) that was calculated as the sum of cytotoxicity and genotoxicity of HAAs following the widely used Tic-Tox approach. Similarly, ML models were developed to predict the HRI, and RF model was found to perform the best, obtaining R2 of 0.69 and RMSE of 0.38. To further explore advanced ML approaches, we developed 3 models using uncertainty-based active learning. Our findings revealed that Categorical Boosting Regression (CAT) model developed through active learning substantially outperformed other models, achieving R2 of 0.87 and 0.82 for predicting concentration and the HRI, respectively. Feature importance analysis with the CAT model revealed that temperature, ions (e.g., chloride and nitrate), and DOC concentration in the distribution network had a significant impact on the occurrence of HAAs. Meanwhile, chloride ion, pH, ORP, and free chlorine were found as the most important features for HRI prediction. This study demonstrates that ML has the potential in the prediction of HAA occurrence and toxicity. By identifying key WQPs impacting HAA occurrence and toxicity, this research offers valuable insights for targeted DBP mitigation strategies.

Hybridizing mechanistic modeling and deep learning for personalized survival prediction after immune checkpoint inhibitor immunotherapy

We present a study where predictive mechanistic modeling is combined with deep learning methods to predict individual patient survival probabilities under immune checkpoint inhibitor (ICI) immunotherapy. This hybrid approach enables prediction based on both measures that are calculable from mechanistic models of key mechanisms underlying ICI therapy that may not be directly measurable in the clinic and easily measurable quantities or patient characteristics that are not always readily incorporated into predictive mechanistic models. A deep learning time-to-event predictive model trained on a hybrid mechanistic + clinical data set from 93 patients achieved higher per-patient predictive accuracy based on event-time concordance, Brier score, and negative binomial log-likelihood-based criteria than when trained on only mechanistic model-derived values or only clinical data. Feature importance analysis revealed that both clinical and model-derived parameters play prominent roles in increasing prediction accuracy, further supporting the advantage of our hybrid approach.

Characterization of cardiac resynchronization therapy response through machine learning and personalized models

Introduction: The characterization and selection of heart failure (HF) patients for cardiac resynchronization therapy (CRT) remain challenging, with around 30% non-responder rate despite following current guidelines. This study aims to propose a novel hybrid approach, integrating machine-learning and personalized models, to identify explainable phenogroups of HF patients and predict their CRT response. Methods:The paper proposes the creation of a complete personalized model population based on preoperative CRT patient strain curves. Based on the parameters and features extracted from these personalized models, phenotypes of patients are identified thanks to a clustering algorithm and a random forest classification is provided. Results:A close match was observed between the 162 experimental and simulated myocardial strain curves, with a mean RMSE of 4.48% (±1.08) for the 162 patients. Five phenogroups of personalized models were identified from the clustering, with response rates ranging from 52% to 94%. The classification results show a mean area under the curves (AUC) of 0.86 ± 0.06 and provided a feature importance analysis with 22 features selected. Results show both regional myocardial contractility (from 22.5% to 33.0%), tissue viability and electrical activation delays importance on CRT response for each HF patient (from 55.8 ms to 88.4 ms). Discussion:The patient-specific model parameters’ analysis provides an explainable interpretation of HF patient phenogroups in relation to physiological mechanisms that seem predictive of the CRT response. These novel combined approaches appear as promising tools to improve understanding of LV mechanical dyssynchrony for HF patient characterization and CRT selection.

Prediction of Hearing Help Seeking to Design a Recommendation Module of an mHealth Hearing App: Intensive Longitudinal Study of Feature Importance Assessment.

Mobile health (mHealth) solutions can improve the quality, accessibility, and equity of health services, fostering early rehabilitation. For individuals with hearing loss, mHealth apps might be designed to support the decision-making processes in auditory diagnostics and provide treatment recommendations to the user (eg, hearing aid need). For some individuals, such an mHealth app might be the first contact with a hearing diagnostic service and should motivate users with hearing loss to seek professional help in a targeted manner. However, personalizing treatment recommendations is only possible by knowing the individual's profile regarding the outcome of interest. This study aims to characterize individuals who are more or less prone to seeking professional help after the repeated use of an app-based hearing test. The goal was to derive relevant hearing-related traits and personality characteristics for personalized treatment recommendations for users of mHealth hearing solutions. In total, 185 (n=106, 57.3% female) nonaided older individuals (mean age 63.8, SD 6.6 y) with subjective hearing loss participated in a mobile study. We collected cross-sectional and longitudinal data on a comprehensive set of 83 hearing-related and psychological measures among those previously found to predict hearing help seeking. Readiness to seek help was assessed as the outcome variable at study end and after 2 months. Participants were classified into help seekers and nonseekers using several supervised machine learning algorithms (random forest, naïve Bayes, and support vector machine). The most relevant features for prediction were identified using feature importance analysis. The algorithms correctly predicted action to seek help at study end in 65.9% (122/185) to 70.3% (130/185) of cases, reaching 74.8% (98/131) classification accuracy at follow-up. Among the most important features for classification beyond hearing performance were the perceived consequences of hearing loss in daily life, attitude toward hearing aids, motivation to seek help, physical health, sensory sensitivity personality trait, neuroticism, and income. This study contributes to the identification of individual characteristics that predict help seeking in older individuals with self-reported hearing loss. Suggestions are made for their implementation in an individual-profiling algorithm and for deriving targeted recommendations in mHealth hearing apps.

Global disparities in CO2 emissions from mobility sectors of diverse economies: A macroscopic exploration across 188 countries/regions

Reducing CO2 emissions represents a global challenge, and the mobility sectors account for a considerable portion of total emissions, with marked disparities across diverse economies. Viewed from a macroscopic perspective, countries and regions around the world can be categorized in various ways. However, relying on a single or a few indicators often proves inadequate to meet the classification requirements for carbon reduction and sustainable development. In this study, employing machine learning and guided by 10 economic indicators, we classified 188 countries/regions into 6 identifiable clusters. Subsequently, by applying ratio analysis and random forest methodologies, we conducted a matrix-based analysis that elucidates the distinct emission characteristics of each mobility sector. Feature importance analysis revealed that for highly developed economies, the total population's contribution was significant, especially in domestic and international aviation, accounting for 50% and 25% of emissions, respectively. In contrast, for lower-middle-income countries and regions, while the total population still played a pivotal role, its influence was most pronounced in railway transportation, reaching 67%. For rapidly developing economies, domestic aviation emissions reached a peak influence of 61%. These insights emphasize the necessity for strategies tailored to the unique attributes of economic entities.

Machine learning-driven feature importance appraisal of seismic parameters on tunnel damage and seismic fragility prediction

This study proposes a machine learning-driven approach for the analysis of the feature importance of seismic parameters on tunnel damage and seismic fragility prediction. The Incremental Dynamic Analysis (IDA) method serves as the fundamental database for vulnerability analysis. Strength and deformation yield criteria are chosen to comprehensively assess the impact of different seismic parameters on the vulnerability of tunnels to seismic events. Three machine learning algorithms, namely Extreme Gradient Boosting (XGBoost), Random Forest (RF), and Support Vector Machine (SVM), are utilized to develop models for classifying and regressing tunnel damage under seismic conditions. Following parameter tuning, the models' performance in multi-classification, binary classification, and regression prediction is assessed, with XGBoost and RF models exhibiting outstanding performance. Feature importance analysis of seismic parameters in XGBoost and RF models for multi-classification, binary classification, and regression is performed using Shapley additive explanations (SHAP). The correlation analysis between SHAP-based feature values and predictions reveals that Peak Ground Displacement (PGD) has the highest influence in the regression model. Utilizing the interaction dependencies among crucial features in the regression model, fragility curves for tunnels based on these key features are effectively derived. The predicted fragility curves closely align with those derived from IDA, illustrating the time-saving and high-performance capabilities of machine learning in nonlinear dynamic computations.

Psychological predictors of online fraud victimhood in China: a machine learning approach

ABSTRACT Understanding why some individuals are more susceptible to becoming victims of fraud is crucial for developing effective anti-fraud strategies. This study employs a machine learning approach to explore the impact of individual psychological and socio-demographic characteristics on susceptibility to fraud. The random forest (RF) models reveal that psychological factors are more influential in determining an individual's vulnerability to fraud than demographic factors. Within the RF models, feature importance analyses highlight that subdimensions of critical thinking – such as truth-seeking, open-mindedness, and cognitive maturity – along with susceptibility to persuasion, perceived benefits on risk, and self-control, are pivotal in influencing an individual’s susceptibility to fraud. These insights are critical for informing targeted interventions and enhancing the effectiveness of anti-fraud measures.

Advancing financial analytics: Integrating XGBoost, LSTM, and Random Forest Algorithms for precision forecasting of corporate financial distress

This study thoroughly examined the use of different machine learning models to predict financial distress in Indonesian companies by utilizing the Financial Ratio dataset collected from the Indonesia Stock Exchange (IDX), which includes financial indicators from various companies across multiple industries spanning a decade. By partitioning the data into training and test sets and utilizing SMOTE and RUS approaches, the issue of class imbalances was effectively managed, guaranteeing the dependability and impartiality of the model’s training and assessment. Creating first models was crucial in establishing a benchmark for performance measurements. Various models, including Decision Trees, XGBoost, Random Forest, LSTM, and Support Vector Machine (SVM) were assessed. The ensemble models, including XGBoost and Random Forest, showed better performance when combined with SMOTE. The findings of this research validate the efficacy of ensemble methods in forecasting financial distress. Specifically, the XGBClassifier and Random Forest Classifier demonstrate dependable and resilient performance. The feature importance analysis revealed the significance of financial indicators. Interest_coverage and operating_margin, for instance, were crucial for the predictive capabilities of the models. Both companies and regulators can utilize the findings of this investigation. To forecast financial distress, the XGB classifier and the Random Forest classifier could be employed. In addition, it is important for them to take into account the interest coverage ratio and operating margin ratio, as these finansial ratios play a critical role in assessing their performance. The findings of this research confirm the effectiveness of ensemble methods in financial distress prediction. The XGBClassifier and RandomForestClassifier demonstrate reliable and robust performance. Feature importance analysis highlights the significance of financial indicators, such as interest coverage ratio and operating margin ratio, which are crucial to the predictive ability of the models. These findings can be utilized by companies and regulators to predict financial distress.

Using machine learning to classify the immunosuppressive activity of per- and polyfluoroalkyl substances

Per- and polyfluoroalkyl substances (PFASs), one of the persistent organic pollutants, have immunosuppressive effects. The evaluation of this effect has been the focus of regulatory toxicology. In this investigation, 146 PFASs (immunosuppressive or nonimmunosuppressive) and corresponding concentration gradients were collected from literature, and their structures were characterized by using Dragon descriptors. Feature importance analysis and stepwise feature elimination are used for feature selection. Three machine learning (ML) methods, namely Random Forest (RF), Extreme Gradient Boosting Machine (XGB), and Categorical Boosting Machine (CB), were utilized for model development. The model interpretability was explored by feature importance analysis and correlation analysis. The findings indicated that the three models developed have exhibited excellent performance. Among them, the best-performing RF model has an average AUC score of 0.9720 for the testing set. The results of the feature importance analysis demonstrated that concentration, SpPosA_X, IVDE, R2s, and SIC2 were the crucial molecular features. Applicability domain analysis was also performed to determine reliable prediction boundaries for the model. In conclusion, this study is the first application of ML models to investigate the immunosuppressive activity of PFASs. The variables used in the models can help understand the mechanism of the immunosuppressive activity of PFASs, allow researchers to more effectively assess the immunosuppressive potential of a large number of PFASs, and thus better guide environmental and health risk assessment efforts.

Target Material Property‐Dependent Cluster Analysis of Inorganic Compounds

The cluster analysis of materials categorizes them according to similarities based on the features of materials, providing insight into the relationship between the materials. Conventional cluster analyses typically use basic features derived from the chemical composition and crystal structure without considering target material properties such as the bandgap and dielectric constant. However, such approaches do not meet demands for grading materials according to properties of interest simultaneously with chemical and structural similarities. Herein, a clustering method grouping similar materials in terms of both the target properties and basic features is proposed. The clustering is compared considering the cohesive energy with that considering the bandgap of metal oxides, showing that their categorizations are clearly different. Further, several clusters classified by the bandgap are analyzed, and coordination environments related to each range of the bandgap are revealed. The clustering for the electronic static dielectric constant identifies a cluster involving several perovskite‐type oxides and balancing with the bandgap near the Pareto front. The method enables analyses with different viewpoints from those of the conventional clustering and feature importance analyses by taking the relationship between the target property and the basic features into account.