eJournal Komunitas Dosen Indonesia
Not a member yet
1406 research outputs found
Sort by
Image Color Correction for Color Vision Deficiency Using ResNet and CycleGAN
Get PDFColor blindness is a visual impairment that limits an individual's ability to accurately perceive certain colors, particularly red, green, or blue. This condition can hinder daily tasks, especially when color identification is crucial. This study proposes a color correction system designed to enhance color perception for individuals with color vision deficiency (CVD), focusing on important visual areas within an image. The method involves converting RGB images into LMS color space, simulating types of color blindness (protanopia, deuteranopia, and tritanopia), detecting visually important regions using a saliency mask, applying color correction through a ResNet-based deep learning model, and performing a reverse transformation back to RGB using a CycleGAN. A total of 5,020 images were used for evaluation, and the proposed system achieved an average Root Mean Square (RMS) error of 0.0212. The Mean Absolute Error (MAE) ranged from 0.1541 to 0.5582 depending on the CVD type. In addition to quantitative evaluation, qualitative validation was conducted through a GUI-based user test involving 10 color blind participants. The system showed the highest effectiveness for deuteranopia with a color recognition accuracy of 71.666%, followed by tritanopia at 59.666% and protanopia at 46.500%. These results indicate that the proposed system offers significant potential in aiding individuals with CVD to better interpret color-based information, especially in visually important regions of an image. Future work may explore broader datasets and alternative deep learning architectures to further improve accuracy and adaptability.Color blindness is a visual impairment that limits an individual's ability to accurately perceive certain colors, particularly red, green, or blue. This condition can hinder daily tasks, especially when color identification is crucial. This study proposes a color correction system designed to enhance color perception for individuals with color vision deficiency (CVD), focusing on important visual areas within an image. The method involves converting RGB images into LMS color space, simulating types of color blindness (protanopia, deuteranopia, and tritanopia), detecting visually important regions using a saliency mask, applying color correction through a ResNet-based deep learning model, and performing a reverse transformation back to RGB using a CycleGAN. A total of 5,020 images were used for evaluation, and the proposed system achieved an average Root Mean Square (RMS) error of 0.0212. The Mean Absolute Error (MAE) ranged from 0.1541 to 0.5582 depending on the CVD type. In addition to quantitative evaluation, qualitative validation was conducted through a GUI-based user test involving 10 color blind participants. The system showed the highest effectiveness for deuteranopia with a color recognition accuracy of 71.666%, followed by tritanopia at 59.666% and protanopia at 46.500%. These results indicate that the proposed system offers significant potential in aiding individuals with CVD to better interpret color-based information, especially in visually important regions of an image. Future work may explore broader datasets and alternative deep learning architectures to further improve accuracy and adaptability
Stacking Ensemble of XGBoost, LightGBM, and CatBoost for Green Economy Index Prediction
No full textIndonesia faces persistent challenges in achieving sustainable development, particularly in harmonizing economic growth with environmental sustainability. The imbalance among economic, social, and environmental dimensions necessitates a comprehensive and reliable measurement tool to assess progress toward a green economy. The Green Economy Index (GEI), developed by the Ministry of National Development Planning (BAPPENAS), serves this function. However, limited data availability at the provincial level, such as in East Java, hampers accurate evaluation and informed policy formulation. This study aims to develop a machine learning-based predictive model for the GEI using a stacking ensemble approach that combines three powerful algorithms: XGBoost, LightGBM, and CatBoost. The model was built using relevant economic, social, and environmental indicators and evaluated on a holdout dataset to assess its predictive accuracy and generalizability. The results show that the stacking ensemble model achieved superior performance compared to the individual models, recording an RMSE of 0.0298, MAE of 0.0225, and the R² score of 0.9774. In comparison, CatBoost, XGBoost, and LightGBM individually performed with slightly lower accuracy. These findings confirm that the stacking ensemble approach is highly effective for predicting GEI values and offers a practical, data-driven solution for supporting sustainable development strategies at the regional level. The study concludes that such predictive tools can significantly enhance policy planning and monitoring of green economic growth, although further research is recommended to validate the model across other provinces.Indonesia faces persistent challenges in achieving sustainable development, particularly in harmonizing economic growth with environmental sustainability. The imbalance among economic, social, and environmental dimensions necessitates a comprehensive and reliable measurement tool to assess progress toward a green economy. The Green Economy Index (GEI), developed by the Ministry of National Development Planning (BAPPENAS), serves this function. However, limited data availability at the provincial level, such as in East Java, hampers accurate evaluation and informed policy formulation. This study aims to develop a machine learning-based predictive model for the GEI using a stacking ensemble approach that combines three powerful algorithms: XGBoost, LightGBM, and CatBoost. The model was built using relevant economic, social, and environmental indicators and evaluated on a holdout dataset to assess its predictive accuracy and generalizability. The results show that the stacking ensemble model achieved superior performance compared to the individual models, recording an RMSE of 0.0298, MAE of 0.0225, and the R² score of 0.9774. In comparison, CatBoost, XGBoost, and LightGBM individually performed with slightly lower accuracy. These findings confirm that the stacking ensemble approach is highly effective for predicting GEI values and offers a practical, data-driven solution for supporting sustainable development strategies at the regional level. The study concludes that such predictive tools can significantly enhance policy planning and monitoring of green economic growth, although further research is recommended to validate the model across other provinces
Convolutional Neural Network Approach for Aspect-Based Sentiment Analysis of Tourism Reviews
Get PDFThe tourism industry is a key economic sector in Indonesia, with East Java ranking highest in tourist visits. This study aims to enhance tourism development by applying aspect-based sentiment analysis (ABSA) using convolutional neural networks (CNN) to analyze online reviews. CNN was selected for this study due to its proven efficiency in capturing local n-gram features and its relatively lower computational cost compared to other deep learning model. Reviews from TripAdvisor and Google Maps were collected focusing on four aspects: attraction, amenities, access, and price. Five different models were developed in this research: one multilabel aspect classifier designed to identify multiple aspects mentioned within each review, and four sentiment classifiers focused on evaluating the sentiment polarity for each specific aspect. These models were trained and evaluated using various combinations of word embeddings, including static embeddings like Word2Vec, and contextualized embeddings such as BERT and IndoBERT. Additionally, the impact of preprocessing through stemming was investigated to understand how simplifying word forms affects model performance. Results indicate that IndoBERT-CNN attains the best overall sentiment classification, reaching F1-scores up to 0.71 for attraction and 0.93 for amenities, while Word2Vec-CNN with stemming leads multilabel classification. Meanwhile stemming improves performance for static embeddings like Word2Vec by simplifying word forms, it reduces effectiveness in transformer-based models like BERT and IndoBERT that rely on natural language context. These findings highlight the benefit of choosing appropriate embeddings and preprocessing for different tasks, thus providing practical insights for improving tourism services through better tourist reviews analysis
Classification of Children's Numerical Intelligence Levels Based on Mathematics Learning Activities Using Machine Learning
Get PDFNumerical ability is a critical aspect of education that impacts both academic achievement and the development of logical thinking skills. However, in Islamic boarding schools (pondok pesantren), the teaching of mathematics often encounters limitations such as a lack of technological integration and diverse learning resources. This study aims to assess students' numerical intelligence levels based on their mathematics learning activities using machine learning techniques, specifically Decision Tree and Random Forest algorithms. Data was collected through questionnaires that captured key variables such as study duration, frequency of practice, learning methods, class engagement, and the speed of problem-solving. After data preprocessing, both models were evaluated through confusion matrices and cross-validation to determine their classification accuracy and stability. The findings revealed that the Decision Tree model achieved an accuracy rate of 86%, while the Random Forest model surpassed it with a 92% accuracy rate, showing more consistent performance. This study highlights the potential of machine learning to enhance educational outcomes, particularly in pesantren settings, by offering deeper insights into students' learning patterns. Beyond classification accuracy, machine learning helps educators identify key factors influencing students' numerical intelligence. The Random Forest model, in particular, revealed that variables such as practice frequency and student engagement are significant predictors of numerical intelligence levels. This information can be used to develop more personalized and effective teaching strategies aimed at improving students' mathematical skills.Numerical ability is a critical aspect of education that impacts both academic achievement and the development of logical thinking skills. However, in Islamic boarding schools (pondok pesantren), the teaching of mathematics often encounters limitations such as a lack of technological integration and diverse learning resources. This study aims to assess students' numerical intelligence levels based on their mathematics learning activities using machine learning techniques, specifically Decision Tree and Random Forest algorithms. Data was collected through questionnaires that captured key variables such as study duration, frequency of practice, learning methods, class engagement, and the speed of problem-solving. After data preprocessing, both models were evaluated through confusion matrices and cross-validation to determine their classification accuracy and stability. The findings revealed that the Decision Tree model achieved an accuracy rate of 86%, while the Random Forest model surpassed it with a 92% accuracy rate, showing more consistent performance. This study highlights the potential of machine learning to enhance educational outcomes, particularly in pesantren settings, by offering deeper insights into students' learning patterns. Beyond classification accuracy, machine learning helps educators identify key factors influencing students' numerical intelligence. The Random Forest model, in particular, revealed that variables such as practice frequency and student engagement are significant predictors of numerical intelligence levels. This information can be used to develop more personalized and effective teaching strategies aimed at improving students' mathematical skills
Prediction of the Impact of Bullying on Students Academic Achievement Using Linear Regression
Get PDFStudent learning achievement is a crucial element in the education sector, shaped by a variety of internal and external factors. Accurately predicting student achievement remains a significant challenge for educators and researchers, especially considering the psychological impacts of bullying. This study aims to construct a predictive model using linear regression to examine the influence of bullying levels, social support, and mental health on student achievement at SMPN 4 Pasarkemis, Tangerang Regency. A review of previous studies highlights the psychological toll bullying can have on academic performance, with many focusing on predictive models or statistical methods like linear regression to quantify these impacts. The model was developed using Python on the Google Colab platform, utilizing the pandas, statsmodels, and seaborn libraries for statistical analysis and visualization of variable relationships. Employing a quantitative, associative research design, the study involved 533 student respondents. The findings reveal that social support has the strongest positive influence on academic achievement, while higher levels of bullying and poorer mental health correlate with decreased performance. Notably, among the various forms of bullying analyzed, cyberbullying emerged as having the most significant negative impact on academic achievement. Although the model explains approximately 5.5 percent of the variation in student learning achievement, the majority of influencing factors lie beyond the scope of this analysis. The model offers potential for further development into a web-based predictive information system to assist educators in the early identification of students at academic risk.Student learning achievement is a crucial element in the education sector, shaped by a variety of internal and external factors. Accurately predicting student achievement remains a significant challenge for educators and researchers, especially considering the psychological impacts of bullying. This study aims to construct a predictive model using linear regression to examine the influence of bullying levels, social support, and mental health on student achievement at SMPN 4 Pasarkemis, Tangerang Regency. A review of previous studies highlights the psychological toll bullying can have on academic performance, with many focusing on predictive models or statistical methods like linear regression to quantify these impacts. The model was developed using Python on the Google Colab platform, utilizing the pandas, statsmodels, and seaborn libraries for statistical analysis and visualization of variable relationships. Employing a quantitative, associative research design, the study involved 533 student respondents. The findings reveal that social support has the strongest positive influence on academic achievement, while higher levels of bullying and poorer mental health correlate with decreased performance. Notably, among the various forms of bullying analyzed, cyberbullying emerged as having the most significant negative impact on academic achievement. Although the model explains approximately 5.5 percent of the variation in student learning achievement, the majority of influencing factors lie beyond the scope of this analysis. The model offers potential for further development into a web-based predictive information system to assist educators in the early identification of students at academic risk
Evaluation of Student UX on E-Learning Web Using User Experience Questionnaire (UEQ)
Get PDFUser Experience (UX) in E-Learning systems plays a critical role in fostering student engagement and improving learning effectiveness. Evaluating UX is essential to address challenges such as reduced learning motivation, poor navigation, and low interface appeal, all of which can hinder educational outcomes. This study aims to evaluate the UX of students using the E-Learning system at SMP Muhammadiyah 17 Surabaya by employing the User Experience Questionnaire (UEQ), which measures six dimensions: Attractiveness, Perspicuity, Efficiency, Dependability, Stimulation, and Novelty. A total of 180 students were selected using Slovin’s formula. Data were analyzed using the UEQ Data Analysis Tool (v10) through descriptive statistics and reliability testing (Cronbach’s alpha > 0.70). The results indicate that the dimensions of Attractiveness (M = 1.43), Efficiency (M = 1.35), Dependability (M = 1.43), and Stimulation (M = 1.34) received high scores and exceeded the average benchmark, indicating that the system is engaging, efficient, reliable, and motivating for users. The Perspicuity dimension (M = 1.42) was rated moderately high, though improvements are needed to better support the understanding of complex materials. In contrast, the Novelty dimension (M = 0.28) scored low, suggesting a lack of innovation and freshness in the interface. These findings highlight the pedagogical significance of UX in enhancing student engagement and online learning effectiveness. Developing more innovative features such as gamified elements or adaptive feedback is recommended to increase the appeal and sustained use of the E-Learning system in alignment with educational goals.User Experience (UX) in E-Learning systems plays a critical role in fostering student engagement and improving learning effectiveness. Evaluating UX is essential to address challenges such as reduced learning motivation, poor navigation, and low interface appeal, all of which can hinder educational outcomes. This study aims to evaluate the UX of students using the E-Learning system at SMP Muhammadiyah 17 Surabaya by employing the User Experience Questionnaire (UEQ), which measures six dimensions: Attractiveness, Perspicuity, Efficiency, Dependability, Stimulation, and Novelty. A total of 180 students were selected using Slovin’s formula. Data were analyzed using the UEQ Data Analysis Tool (v10) through descriptive statistics and reliability testing (Cronbach’s alpha > 0.70). The results indicate that the dimensions of Attractiveness (M = 1.43), Efficiency (M = 1.35), Dependability (M = 1.43), and Stimulation (M = 1.34) received high scores and exceeded the average benchmark, indicating that the system is engaging, efficient, reliable, and motivating for users. The Perspicuity dimension (M = 1.42) was rated moderately high, though improvements are needed to better support the understanding of complex materials. In contrast, the Novelty dimension (M = 0.28) scored low, suggesting a lack of innovation and freshness in the interface. These findings highlight the pedagogical significance of UX in enhancing student engagement and online learning effectiveness. Developing more innovative features such as gamified elements or adaptive feedback is recommended to increase the appeal and sustained use of the E-Learning system in alignment with educational goals
Development of an Android Application for Recipe Management Using Flutter and API
Get PDFThe advancement of digital technology encourages micro, small, and medium enterprises (MSMEs) to adapt their operational systems, including food recipe management. Many culinary businesses still rely on manual methods, which are prone to data loss, ingredient inconsistency, and inefficiencies. This condition often leads to increased operational costs, food waste, and inconsistency in product quality, which ultimately affects customer satisfaction and business sustainability. Digital solutions, especially those that support real-time recipe standardization, have the potential to streamline operations, minimize errors, and improve production consistency. This study aims to design and develop an Android-based food recipe management application using Flutter and REST API to support the digital transformation of MSMEs in the culinary sector. The Research and Development (R&D) approach with a prototyping model was applied, involving iterative stages: literature review, needs analysis, interface design using Figma, implementation using Flutter and REST API, and Black Box Testing involving real users for functional evaluation. The results show that all core features login, recipe search, categorization, and recipe data management (add, edit, delete) functioned properly as intended. User feedback indicated increased operational efficiency, reduced manual effort, and improved consistency in recipe handling, particularly in onboarding new staff and ensuring product uniformity. In conclusion, the developed application contributes practically to improving recipe management efficiency and supporting the digital transformation of MSMEs. This system also lays the groundwork for future development of AI-powered features, automatic nutrition analysis, and cross-platform expansion to iOS and web
Analysis of Factors Affecting the Net Asset Value of Sharia Equity Mutual Funds
Get PDFOne type of collective investment vehicle that manages public monies for investment is a sharia stock mutual fund stocks that are in accordance with sharia law and avoid transactions that contain elements of usury, maisir, and gharar. This study aims to investigate how the money supply, inflation, gold prices, and age of mutual funds impact the performance of Indonesian mutual funds' net worths that invest in sharia stocks between 2019 and 2024. This study made use of secondary data study. The sample used in this study was 6 sharia stock mutual fund companies, using purposive sampling techniques and there was a population of 55 sharia stock mutual funds. Using Eviews 12 sv software, utilizing the panel data regression approach, was used to secondary data. The findings of the study indicate that the quantity of currency in use, the cost of gold, and the age of mutual funds all have a major influence on the value of net assets (NAV) of Indonesian Sharia Equity Mutual Funds, however inflation has no discernible effect on this number
Pengaruh Work Life Balance, Lingkungan Kerja, dan Kedisiplinan Kerja Terhadap Kepuasan Kerja Karyawan CV. FA Chemical
Get PDFRiset ini dimaksudkan untuk menganalisis pengaruh Work Life Balance, Lingkungan Kerja, dan Kedisiplinan Kerja pada Kepuasan Kerja karyawan pada CV. FA Chemical. Pendekatan kuantitatif dengan teknik sampel jenuh yang melibatkan seluruh populasi, yaitu 79 karyawan sebagai responden. Pengumpulan data melalui angket dan dianalisis dengan SPSS mempergunakan uji, seperti uji validitas, reliabilitas, regresi linear berganda, uji t, uji f, serta uji asumsi klasik. Work Life Balance dan Lingkungan Kerja berkontribusi positif dan signifikan pada Kepuasan Kerja, sementara Kedisiplinan Kerja tidak berkontribusi signifikan. Secara simultan, ketiga variabel memberikan kontribusi sebesar 86% pada variasi Kepuasan Kerja karyawan. Nilai R sebesar 0,927 menunjukkan hubungan yang sangat kuat antara ketiga variabel independen dengan kepuasan kerja. Selain itu, nilai Durbin-Watson sebesar 1,934 menandakan tidak adanya autokorelasi, sehingga model regresi layak dipergunakan. Keseimbangan yang baik antara kehidupan pribadi dan pekerjaan serta lingkungan kerja yang nyaman menjadi faktor utama dalam meningkatkan kepuasan kerja karyawan. Di sisi lain, meskipun kedisiplinan diperlukan untuk kelancaran operasional perusahaan, pengaruhnya terhadap tingkat kepuasan kerja tidak signifikan. Oleh karena itu, perusahaan disarankan untuk lebih mengutamakan peningkatan Work Life Balance dan menciptakan lingkungan kerja yang mendukung guna meningkatkan kepuasan serta produktivitas karyawan secara menyeluruh
Pengaruh Profitabilitas, Likuiditas, Leverage, dan Ukuran Perusahaan Terhadap Tax Avoidance
Get PDFPenelitian ini bertujuan untuk mengkaji bagaimana penghindaran pajak dipengaruhi oleh leverage, ukuran perusahaan, profitabilitas, dan likuiditas. Perusahaan properti dan real estate yang terdaftar di Bursa Efek Indonesia tahun 2021 - 2024 menjadi objek penelitian ini. Purposive sampling diterapkan untuk memilih sampel. Sampel terdiri dari total 104 perusahaan. Hasil penelitian ini dianalisis menggunakan perangkat lunak SPSS, dan analisis regresi linier berganda digunakan untuk menganalisis. Hasil penelitian menunjukkan bahwa Profitabilitas dihitung menggunakan Return On Assets (ROA) dan ukuran perusahaan (SIZE) diukur dari logaritma dari total aset memiliki pengaruh negatif dan signifikan kepada CETR, yang menunjukkan jika perusahaan yang menghasilkan keuntungan dan berukuran besar cenderung melakukan tax avoidance. Leverage diukur menggunakan Debt to Equity Ratio (DER) memiliki pengaruh positif dan signifikan kepada CETR, yang berarti bahwa company yang memiliki tingkat leverage tinggi condong tidak melaksanakan tax avoidance. Sementara itu, likuiditas di ukur menggunakan Current Ratio (CR) tidak berpengaruh signifikan terhadap CETR, sehingga likuiditas tidak berkaitan secara langsung dengan tax avoidance