Sample Selection Method Research Articles

Integrating a multi-dimensional deep convolutional neural network with optimized sample selection for landslide susceptibility assessment

ABSTRACT To address the errors of negative samples in landslide susceptibility modeling and traditional methods in exploring the regularities hidden in the evaluation factors, this paper proposes a stacking one- and three-dimensional Convolutional Neural Network (Stacking-1D-3D-CNN) landslide susceptibility assessment method considering sample optimization selection. First, in order to select negative samples rationally, this paper adopts the Relative Frequency Ratio combined with Certainty Factor Method (RFR-CFM) to determine the negative samples; secondly, the Stacking-1D-3D-CNN proposed is combined with RFR-CFM for the first time for landslide susceptibility assessment. In this work, the negative samples determined by RFR-CFM and the Information Quality Model (IQM) were combined with historical disaster points to form a total modeling sample, and modeled at different ratios. Finally, it is compared with several other models in terms of the landslide hazard susceptibility zoning results, prone zone statistics, and model performance. The findings show that the degree of spatial aggregation of training samples and testing samples has a much greater impact on the accuracy of landslide susceptibility modeling than the impact of their proportions. Furthermore, compared with other models, RFR-CFM-Stacking-1D-3D-CNN has the highest AUC value, precision, recall, F-score, and accuracy, which are 0.95, 0.83, 0.89, 0.85, and 84.76%, respectively, and the lowest RMSE and MAE, 0.39 and 0.15, respectively. This proves the RFR-CFM sample selection method’s rationality and the Stacking-1D-3D-CNN model’s effectiveness.

Sample selection using multi-task autoencoders in federated learning with non-IID data

Federated learning is a machine learning paradigm in which multiple devices collaboratively train a model under the supervision of a central server while ensuring data privacy. However, its performance is often hindered by redundant, malicious, or abnormal samples, leading to model degradation and inefficiency. To overcome these issues, we propose novel sample selection methods for image classification, employing a multi-task autoencoder to estimate sample contributions through loss and feature analysis. Our approach incorporates unsupervised outlier detection, using one-class support vector machine (OCSVM), isolation forest (IF), and adaptive loss threshold (AT) methods managed by a central server to filter noisy samples on clients. We also propose a multi-class deep support vector data description (SVDD) loss controlled by a central server to enhance feature-based sample selection. We validate our methods on CIFAR10 and MNIST datasets across varying numbers of clients, non-IID distributions, and noise levels up to 40%. The results show significant accuracy improvements with loss-based sample selection, achieving gains of up to 7.02% on CIFAR10 with OCSVM and 1.83% on MNIST with AT. Additionally, our federated SVDD loss further improves feature-based sample selection, yielding accuracy gains of up to 0.99% on CIFAR10 with OCSVM. These results show the effectiveness of our methods in improving model accuracy across various client counts and noise conditions.

APPLICATION OF GUIDED INQUIRY MODEL ASSISTED BY AUDIO-VISUAL MEDIA TO IMPROVE INQUIRY SKILLS IN SCIENCE LEARNING

This study examines the improvement of students' inquiry skills facilitated through the implementation of a guided inquiry learning strategy accompanied by the use of audiovisual media, and measures the magnitude of the improvement of these skills after the implementation of this strategy. This study adopted a nonequivalent control group design as a type of quasi-experiment. All students in grade VIII at SMP Negeri 1 Garung became the target population. The selected sample subjects were VIII C class as the treatment group, and VIII A class as the comparison group with non-random sample selection method. The instruments utilized in this research are test devices, observation sheets, and documentation. The results of the analysis showed that the inquiry skills of students who received a guided inquiry strategy accompanied by audiovisual media had a higher average than those who only received a guided inquiry strategy without audiovisual media. The results of the comparison test of the means of two independent groups with a significance level of 5% resulted in a significance value of 0.000 <0.05, so ha is accepted while ho is rejected. Testing with the N-Gain metric showed that the average score of the treatment group was 0.75, which was in the high category, while the average score of the comparison group was 0.68, which was in the medium category. Based on this analysis, it can be seen that the guided inquiry strategy with audiovisual media support is effective in improving students' inquiry skills.

Awareness and Determination of Digital Game Addiction Levels among Nursing Students

Objective: This study aimed to determine nursing students’ awareness of digital game addiction and their addiction levels. Method: This cross-sectional study was conducted between 01.12.2022 and 01.03.2023. The study’s universe was the nursing department of a university. The sample size was calculated as n = 362 using the known sample selection method. Data were collected using a personal information form, the Digital Game Addiction Scale for University Students, and the Digital Game Addiction Awareness Scale. The Mann–Whitney U test, the Kruskal–Wallis test, Spearman’s rho, and logistic regression were used in the analyses. Results: It was determined that males (7.136 times), those with psychiatric illnesses (0.137 times), those who received education on digital games (0.278 times), and those who used their phone for extended periods (1.228 times) had a higher risk of digital game addiction (p < 0.05). When the factors related to digital game addiction awareness were examined, women (0.522 times), those with a high income (2.894 times), ad those with low digital game addiction scores (0.901) were less at risk (p < 0.05). Conclusion: Digital game addiction scores increase as phone-computer usage and phone game playing time increase. In terms of digital addiction, males, those with a history of psychiatric illness, those who received education through digital games, and those who used their phones for long periods were at a higher risk. Women, those with a high-income status, those with low digital game addiction scores, and those with a heightened awareness of digital game addiction are more aware of digital game addiction.

The effect of leverage and capital intensity on tax avoidance with institutional ownership as a moderating variable

This research is a quantitative study that aims to analyze the effect of leverage and capital intensity on tax avoidance with institutional ownership as a moderating variable. This research is a quantitative study using secondary data in the form of audited financial reports on Primary Consumer Goods sector companies listed on the Indonesia Stock Exchange for the period 2019 - 2023. The sample selection method used purposive sampling technique with several criteria and 304 samples were used in this study. The data analysis used in this study is descriptive statistical analysis, parameter significance test (t test), and determination coefficient test using STATA v.17.0 software with a significance level of 5% (0.05). The results of this test obtained (1) there is no significant effect of leverage on tax avoidance, (2) there is no significant effect of capital intensity on tax avoidance, (3) institutional ownership is not able to moderate the effect of leverage on tax avoidance, (4) institutional ownership is not able to moderate the effect of capital intensity on tax avoidance.

HUBUNGAN ASUPAN MAKAN, PERSEPSI CITRA TUBUH, DAN PERILAKU PENGENDALIAN BERAT BADAN DENGAN STATUS GIZI GURU DI PERGURUAN MUHAMMADIYAH RAWAMANGUN

Overnutrition is one of the most common nutritional problems in the adult age group. Overnutrition among working adults can harm productivity at work. This is because excess nutritional status can cause various non-communicable diseases such as type 2 diabetes mellitus, coronary heart disease, and cancer. The purpose of this study was to determine the relationship between food intake, body image perception, and weight control behavior with nutritional status at the Muhammadiyah College of Rawamangun. The research was conducted in August 2023, involving 61 teachers. The research method used is quantitative research with a cross-sectional design and total sampling method for sample selection. Data was analyzed using the Chi-Square statistical test. Dietary intake was measured using a 2x24-hour recall method, body image perception and weight control behavior were assessed using Body Shape Questionnaire (BSQ) Short Version also waist circumference, body weight and height were measured using anthropometric measurements. The results of data analysis indicate a significant relationship between energy intake, protein intake, fat intake, carbohydrate intake, body image perception, and weight control behavior with nutritional status based on BMI and waist circumference (P<0.05). However, no significant relationship was found between body image perception and weight control behavior with nutritional status based on waist circumference (p>0.05). In conclusion, there were significant relationship between energy intake, protein intake, fat intake, carbohydrate intake, body image perception, and weight control behavior with nutritional status based on BMI and waist circumference.

Discovering Representations of Democracy in Big Data: Purposive Semantic Sample Selection for Qualitative and Mixed-Methods Research

The increasing volume of large, multi-thematic text corpora in social sciences presents a challenge in selecting relevant documents for qualitative and mixed-methods research. Traditional sample selection methods require extensive manual coding or prior dataset knowledge, while unsupervised methods can yield inconsistent results with theory-driven coding. To address this, we propose purposive semantic sampling – a Natural Language Processing approach using document-level embeddings created by a weighted average of word vectors with term frequency-inverse document frequency (tf-idf). We demonstrate its effectiveness using the example of democracy, a complex topic difficult to retrieve from parliamentary corpora. This method applies to any multi-thematic research area within big data, offering a reliable, efficient sample selection method for social research texts. Our contribution includes validating this NLP approach for social sciences and humanities as well as providing a robust tool for researchers, facilitating deeper qualitative analysis and exploration of big data corpora within the computational grounded theory framework.

Pengaruh Pemberian Mata Kuliah Perpajakan, Pelatihan Brevet Pajak dan Pertimbangan Pasar Kerja Terhadap Minat Mahasiswa Berkarir di Bidang Perpajakan

The purpose of this study is to analyze the influence of the Provision of Tax Courses, Tax Brevet Training and Job Market Considerations on Student Interest in Career in the Field of Taxation. This type of research is a quantitative research with primary data sources. The population in this study is Pamulang University students with the Accounting Study Program and have or are taking Brevet. The sample selection method used in this study is purposive sampling so that the final sample obtained is 80 students. Testing the data quality using validity and reliability tests using IBM SPSS25 software, the next step is to test the classic assumptions, namely the normality test, the multicollinearity test and the hexaoscedasticity test after that the hypothesis test. From the results that have been carried out in this study, it shows that Taxation Courses, Tax Brevet Training and Job Market Considerations simultaneously affect the Interest of Students in Career in the Field of Taxation. Partially shows that the Taxation and Market Consideration Course has an effect on the Interest of Career Students in the Field of Taxation. Meanwhile, the Tax Brevet Training partially has no effect on the interest of students in a career in taxation in students of the University of Pamulang Accounting Study Program.

PENGARUH RELIGIUSITAS, GENDER, DAN LOVE OF MONEY TERHADAP PERSEPSI MAHASISWA TERKAIT PENGGELAPAN PAJAK

The purpose of this study is to analyze the influence of Religiosity, Gender and Love of Money on Student Perceptions Related to Tax Evasion. This type of research is a quantitative research with primary data sources. The population in this study is Pamulang University students with the Accounting Study Program with a Concentration in Taxation. The sample selection method used in this study was convenience sampling so that the final sample obtained was 94 students and rounded to 100 students. The data quality test was carried out using a validity and reliability test and using IBM SPSS25, after which a classical assumption test was carried out, namely a normality test, a multicollinearita test and a heteroscedasticity test, after which a hypothesis test was carried out. The results of this study show that Religiosity, Gender and Love of Money simultaneously affect Student Perception Related to Tax Evasion. It partially shows that Love of Money has an effect on Student Perception Related to Tax Evasion. As for Religiosity and Gender, it does not affect Student Perception Related to Tax Evasion carried out on students of Pamulang University Accounting Study Program with Tax Concentration. Keywords: Student Perception Related to Tax Evasion, Religiosity, Gender, Love of Money

The Influence of Blue Accounting, Corporate Social Responsibility, and Stakeholders Influence Capacity on Financial Performance

This study aims to determine the effect of green accounting, blue accounting, corporate social responsibility (CSR), and stakeholders influence capacity on financial performance. The sample in this study was the Maritime sector companies listed on the Indonesia Stock Exchange in 2021-2022. In drawing samples, the author used a purposive sampling technique, namely a sample selection method based on certain criteria. The data used were obtained from the Indonesia Stock Exchange publication and the official website of each company. This study used 24 samples with 2 years of observation so that the total was 48 observations with multiple linear regression estimation of panel data. The results showed that the green accounting variable was negative and insignificant on financial performance, blue accounting had a positive and insignificant effect on financial performance, corporate social responsibility had a positive and insignificant effect on financial performance, stakeholders influence capacity had a positive and significant effect on financial performance.

Vulnerability Analysis: A Tool for SMEs’ Business Resilience

This study is aimed at examining the vulnerability levels of businesses and assessing the influence of vulnerability on business resilience. This study was undertaken in Surakarta, Central Java, Indonesia. The sample selection method utilizes purposive random sampling. Data is gathered through structured interviews. The vulnerability level of businesses is gauged using an index. The vulnerability index was formulated based on three pivotal variables: exposure, sensitivity, and adaptive capacity. The vulnerability index was ascertained through the aggregation of values derived from all vulnerability indicators. The impact of vulnerability on changes in the external business environment resilience is analyzed through multinomial logistic regression. The research outcomes indicate that SMEs in the food and beverage sector fall within the bracket of moderate vulnerability. Sensitivity emerges as the most pivotal factor influencing the vulnerability index of business units, closely followed by adaptive capacity and exposure. These variables demonstrate a robust impact on SMEs resilience. The critical need for both government and business manager to prioritize ensuring access to suppliers and raw materials, manage raw material pricing, optimize product pricing strategies, and maintain market reach after macro-environmental uncertainty (disaster). This study emphasizes the importance of the durability of SMEs the nimble external environment, where access to suppliers and the price of raw materials, product prices, and market reach often change. It also reveals that sensitivity is a vital variable that affects the durability of SMEs.

The Role of Financial Rations on Financial Distress Conditions in Transportation and Logistics Sector Companies at Indonesian Sharia Stock Index

Financial distress as an early warning signal of conditions that can lead to bankruptcy of a company. This study aims to examine the effect of financial ratios, namely liquidity, profitability, solvency, and activity on the financial distress of a company. The type of data used in this study is secondary data sourced from financial reports available at www.idx.com. The sample selection method is carried out by purposive sampling. The sample in this study was 44 Transportation and Logistics companies listed on the Indonesian Sharia Stock Index during the 2020-2023 period. The data testing tool used is multiple linear regression. The results of the F test analysis show that simultaneously, the variables of liquidity, profitability, solvency and activity have a significant effect on financial distress. Meanwhile, the results of the T test show that the variables of liquidity, profitability, solvency and activity have a significant effect on financial distress.

Replay-Based Incremental Learning Framework for Gesture Recognition Overcoming the Time-Varying Characteristics of sEMG Signals.

Gesture recognition techniques based on surface electromyography (sEMG) signals face instability problems caused by electrode displacement and the time-varying characteristics of the signals in cross-time applications. This study proposes an incremental learning framework based on densely connected convolutional networks (DenseNet) to capture non-synchronous data features and overcome catastrophic forgetting by constructing replay datasets that store data with different time spans and jointly participate in model training. The results show that, after multiple increments, the framework achieves an average recognition rate of 96.5% from eight subjects, which is significantly better than that of cross-day analysis. The density-based spatial clustering of applications with noise (DBSCAN) algorithm is utilized to select representative samples to update the replayed dataset, achieving a 93.7% recognition rate with fewer samples, which is better than the other three conventional sample selection methods. In addition, a comparison of full dataset training with incremental learning training demonstrates that the framework improves the recognition rate by nearly 1%, exhibits better recognition performance, significantly shortens the training time, reduces the cost of model updating and iteration, and is more suitable for practical applications. This study also investigates the use of the incremental learning of action classes, achieving an average recognition rate of 88.6%, which facilitates the supplementation of action types according to the demand, and further improves the application value of the action pattern recognition technology based on sEMG signals.

The Impact of Training Distances (20 m, 50 m, and 70 m) on Concentration and Anaerobic Endurance in Archery Sub-Elite Athletes

This study aims to investigate the effects of a new training method for archery athletes with different distances, namely 20 meters, 50 meters, and 70 meters, in improving concentration and anaerobic endurance. The pre- and post-control group design was used in this study with the sample selection method using the purposive sampling technique. The criteria for participants in this study include: actively practicing archery, aged 14-20 years, having at least 5 years of archery competition experience, domiciled in East Java Province, not injured, and willing to participate in the study until the end. 40 sub-elite archery athletes (20 males, 20 females) became participants in this study and were divided into 4 groups: conventional training (CON), training with a shooting distance of 20 m (NEAR), training with a shooting distance of 50 m (MID), and training with a shooting distance of 70 m (LONG) training groups. Each group underwent a 12-week training program specific to their assigned distance. The instrument used to measure concentration is the standardised archery mirror drawing test and to measure anaerobic endurance using the RAST test instrument. The results of the paired sample t-test analysis showed that in the anaerobic endurance variable, the NEAR, MID, and LONG groups had a p-value <0.05, while the CON group had a p-value > 0.05. In the concentration variable (time), all groups had a p-value <0.05, while in the concentration variable (error), there were only two groups (NEAR and CON) that had a p-value <0.05. The Anova Test results showed a significant difference between the CON and MID groups and CON and LONG (p <0.05) in the anaerobic endurance, while in the concentration (time) there was a significant difference between the CON and LONG groups and NEAR and LONG (p<0.05). In conclusion, the MID and LONG groups significantly improve anaerobic endurance and concentration, two important factors for archery success. Longer training distances appear to have a more pronounced effect on concentration.

Pengaruh Solvabilitas, Profitabilitas, Kepemilikan Institusional, Dewan Komisaris Independen, dan Kepemilikan Publik terhadap Pengungkapan Manajemen Risiko: Studi pada Perusahaan Sub-sektor Perbankan yang Terdaftar di Bursa Efek Indonesia Tahun 2018-2022

The object of this research is a banking sub-sector company listed on the Indonesia Stock Exchange for the period 2018-2022. This study aims to determine the effect of leevrage, profitability, institutional ownership, independent board of commissioners, and public ownership as independent variables and risk management disclosure as the dependent variable. This study used 35 companies, so there were a total of 175 samples used. The study used purposive sampling as a sample selection method used for the selection of certain criteria. This study applied descriptive statistical analysis method and panel data regression analysis. The results of this study show that all variables simultaneously influence risk management disclosure. Meanwhile, leverage, profitability, and independent board of commissioners each have no effect. Meanwhile, institutional ownership and public ownership positively affect risk management disclosure.

Education and women in high-skill occupations in India

Purpose This paper aims to examine women’s likelihood of being employed in a high-skill occupation compared to men in India. It also analyses the impact of education on women’s likelihood of entering a high-skill occupation. Design/methodology/approach This study uses data from Periodic Labour Force Survey 2018–19. National Classification of Occupations 2004 is used to classify the occupations in three categories, i.e. high-skill, medium-skill and low-skill. The paper uses ordered probit with Heckman’s sample selection method to estimate the impact of gender identity and education on women’s likelihood of entering high-skill occupations. Findings The results show that female workers are more likely to be employed in high-skill occupations than male workers. Formal years of schooling and technical education further improve women’s probability of working in high-skill occupations compared to men. However, an increase in monthly per capita expenditure of the households improves men’s chances of being employed in high-skill occupations in comparison to women. It suggests that gender norms are still relevant. Originality/value Studies have found occupational segregation in India with a large share of women being employed in low-wage work. This study examines the issue by focusing on the skill level of occupations. It found that women have a higher likelihood of entering high-skill occupations in India and education is playing an important role in it.

Pengaruh Return On Asset, Free Cash Flow, Firm Size, Terhadap Dividend Payout Ratio Dengan Corporate Governance Sebagai Variabel Intervening Pada Perusahaan Manufaktur

This research aims to determine Return on Asset, Free Cash Flow, and Firm Size on Dividend Payout Ratio with Corporate Governance as an intervening variable in manufacturing companies listed on the Indonesia Stock Exchange for the period 2017 - 2021. The sample selection method used is purposive sampling. The total observations in this research were 155 observations consisting of 31 companies. The data analysis technique used in this research is path analysis with the SPSS version 26 analysis tool. The test results of sub structure I show that Return on Asset and Free Cash Flow have a positive effect on Corporate Governance. While Firm Size has no effect on Corporate Governance. The results of sub-structure II testing show that Return on Asset and Free Cash Flow have a positive effect on Dividend Payout Ratio. Meanwhile, Corporate Governance have a negative effect on Dividend Payout Ratio and Firm Size has no effect on Dividend Payout Ratio. The sobel test results show that Corporate Governance is able to mediate the effect of Return on Asset and Free Cash Flow on Dividend Payout Ratio. Meanwhile, Corporate Governance is unable to mediate the effect of Firm Size on the Dividend Payout Ratio.

Pengaruh Partisipasi Anggaran dan Kejelasan Anggaran terhadap Kesenjangan Anggaran dengan Job Relevant Information sebagai variable moderating

This study aims to test and analyze the effect of budget participation, clarity of budget targets on budgetary slack with Job Relevant Information as a moderating variable. This study was conducted at PT. Indomobil Prima Niaga Sumut Aceh (consisting of 5 branches). The sample selection method of saturated sampling is a sampling technique when all members of the population are used as samples, namely 38 samples. The sample in this study was the Head of the department consisting of Finance and Accounting Head, Aftersales Head, Branch Manager, Regional Manager and Hoyu Manager who are involved in budget preparation, budget implementation and budget accountability. The research approach used in this study is associative research. This type of research is quantitative research. Data collection was carried out using a survey method by distributing questionnaires to respondents. The analysis technique used is the Quantitative Approach Using Statistical Analysis, namely Partial Least Square - Structural Equestion Model (PLS SEM). The results of the study indicate that budget participation has an effect on budgetary slack. Furthermore, clarity of budget targets has an effect on budgetary slack. Job relevant information moderates the effect of budget participation and clarity of budget targets on budgetary slack

Label-noise learning via uncertainty-aware neighborhood sample selection

Existing deep learning methods often require a large amount of high-quality labeled data. Yet, the presence of noisy labels in the real-world training data seriously affects the generalization ability of the model. Sample selection techniques, the current dominant approach to mitigating the effects of noisy labels on models, use the consistency of sample predictions and observed labels to make clean selections. However, these methods rely heavily on the accuracy of the sample predictions and inevitably suffer when the model predictions are unstable. To address these issues, we propose an uncertainty-aware neighborhood sample selection method. Especially, it calibrates for sample prediction by neighbor prediction and reassigns model attention to the selected samples based on sample uncertainty. By alleviating the influence of prediction bias on sample selection and avoiding the occurrence of prediction bias, our proposed method achieves excellent performance in extensive experiments. In particular, we achieved an average of 5% improvement in asymmetric noise scenarios.

Pluto: Sample Selection for Robust Anomaly Detection on Polluted Log Data

Log anomaly detection, critical in identifying system failures and preempting security breaches, finds irregular patterns within large volumes of log data. Modern log anomaly detectors rely on training deep learning models on clean anomaly-free log data. However, such clean log data requires expensive and tedious human labeling. In this paper, we thus propose a robust log anomaly detection framework, PlutoNOSPACE, that automatically selects a clean representative sample subset of the polluted log sequence data to train a Transformer-based anomaly detection model. Pluto features three innovations. First, due to localized concentrations of anomalies inherent in the embedding space of log data, Pluto partitions the sequence embedding space generated by the model into regions that then allow it to identify and discard regions that are highly polluted by our pollution level estimation scheme, based on our pollution quantification via Gaussian mixture modeling. Second, for the remaining more slightly polluted regions, we select samples that maximally purify the eigenvector spectrum, which can be transformed into the NP-hard facility location problem; allowing us to leverage its greedy solution with a (1-(1/e)) approximation guarantee in optimality. Third, by iteratively alternating between the above subset selection, a model re-training on the latest subset, and a subset filtering using dynamic training artifacts generated by the latest model, the data selected is progressively refined. The final sample set is used to retrain the final anomaly detection model. Our experiments on four real-world log benchmark datasets demonstrate that by retaining 77.7% (BGL) to 96.6% (ThunderBird) of the normal sequences while effectively removing 90.3% (BGL) to 100.0% (ThunderBird, HDFS) of the anomalies, Pluto provides a significant absolute F-1 improvement up to 68.86% (2.16% → 71.02%) compared to the state-of-the-art sample selection methods. The implementation of this work is available at https://github.com/LeiMa0324/Pluto-SIGMOD25.