Fine-grained recognition of cultural artifacts remains challenging because of the scarcity of annotated data, subtle intra-class differences, and heterogeneous imaging conditions. This study addresses these issues through a domain-specific deep learning pipeline, demonstrated on Indonesian keris classification across three tasks: pamor (27 classes), dhapur (42), and tangguh (5). The pipeline integrates background homogenization, orientation normalization, and YOLOv8-based blade cropping with mask-aware augmentation restricted to the blade regions. For classification, we propose KerisRDNet, which extends InceptionResNetV2 with Inception-Residual-Dilated (IRD) blocks and squeeze-and-excitation to model the elongated geometries and subtle forging motifs. Experiments show that baseline networks collapse under fine-grained settings, with macro-F1 near zero, whereas the proposed approach achieves 0.268 (pamor), 0.276 (dhapur), and 0.635 (tangguh) with Top-3 accuracy above 0.5 and AUC up to 0.853. Across three stratified resamplings, paired non-parametric tests (Wilcoxon signed-rank) indicated directionally consistent improvements; given the small number of repetitions (n=3), these results are interpreted conservatively. These results demonstrate the feasibility of practically viable keris recognition as a decision-support tool for cultural heritage curation, while also offering a transferable workflow for low-data fine-grained recognition tasks.
The rapid growth of e-commerce in Indonesia increases the need for sentiment analysis to accurately understand customer perceptions. This study evaluates the effectiveness of the Transformer-based IndoBERT model for sentiment classification on Indonesian e-commerce reviews and compares its performance with four RNN architectures (LSTM, GRU, BiLSTM, and BiGRU). The PRDECT-ID dataset containing 5,400 reviews was processed through preprocessing, an 80:20 data split, RNN training using 5-Fold Cross Validation, and IndoBERT fine-tuning under a hold-out scheme. Unlike previous studies that focused solely on RNN models with a maximum accuracy of 90.7%, this work expands the evaluation by integrating a Transformer-based approach. Results show that IndoBERT achieves 98.52% accuracy and F1-weighted score, outperforming the best RNN models by approximately 0.94–0.95. Paired T-Test and Wilcoxon tests yield p < 0,05, confirming that the performance improvements are statistically significant. IndoBERT demonstrates greater stability and effectiveness for Indonesian sentiment analysis.
This study aims to analyze the effect of Financial Literacy and Fintech utilization on Green Investment Intention, as well as to examine the role of Risk Tolerance as a moderating variable among the general public in Indonesia. The study uses a quantitative approach with an online survey of 150 respondents aged at least 18 years who have access to digital financial services and an interest in sustainable investment, using purposive sampling. Data analysis was conducted using the Partial Least Squares–Structural Equation Modeling (PLS-SEM) method. The results show that Financial Literacy and the use of Fintech have a positive effect on Green Investment Intention, while Risk Tolerance has a direct effect and strengthens the relationship between Financial Literacy and Fintech with Green Investment Intention. The research model was able to explain 70.2% of the variation in Green Investment Intention, confirming the importance of cognitive, technological, and psychological factors in encouraging sustainable investment. The novelty of this research lies in the integration of financial literacy, Fintech, and Risk Tolerance as moderating variables in the sustainable financial behavior model among the general public in Indonesia. These findings contribute to the development of sustainable finance literature and provide practical implications for governments, financial institutions, and Fintech providers in designing financial education policies and digital services that support the transition to a green economy.
Latar belakang: Perempuan menjalankan peran domestik sebagai aktivitas rutin yang mengandung risiko bahaya ergonomi. Aktivitas domestik melibatkan kerja pengangkatan beban, postur, gerak berulang dan statis. Tujuan: untuk mengidentifikasi prevalensi nyeri musculoskeletal, aktivitas domestik dominan dan memberikan edukasi terkait bahaya ergonomis yang muncul dari aktivitas domestik. Metode: Kegiatan ini dilaksanakan dengan pendekatan survei dan edukasi. Identifikasi nyeri menggunakan kuesioner pemetaan lokasi nyeri tubuh dan Visual Analogue Scale (VAS). Edukasi dilakukan melalui penyuluhan interaktif, diskusi, tanya jawab, dan pemberian contoh penerapan prinsip ergonomi pada aktivitas domestik di kelompok masyarakat Dawis Kenanga, dan dianalisis secara deskriptif statistik. Hasil: Tingkat nyeri yang dirasakan memiliki nilai rerata dada sebesar rerata 4,17 + 2,43 dengan area tubuh nyeri dominan pada dada, bahu, lutut, leher, punggung, pinggul. Aktivitas dominan yang menyebabkan keluhan tertinggi yakni aktivitas mengurus pakaian (52,94%) seperti menyetrika, mencuci, menjemur. Kesimpulan: Pelaksanaan edukasi terhadap bahaya ergonomi dan pengendalian risiko pada aktivitas domestik meningkatkan pemahaman 88,2% partisipan memberikan nilai dampak sebesar 0,91.
Retail transaction data can be utilized to identify consumer purchasing patterns and product relationships. This study evaluates and compares the performance of the Apriori and ECLAT algorithms in generating association rules from a retail dataset comprising 30,000 transactions collected between January and April 2025. A grid search method was employed to determine the optimal minimum support and minimum confidence thresholds. Using a minimum support of 0.005 and a minimum confidence of 0.3, both algorithms generated 1,736 frequent itemsets and 78 association rules. The resulting rules were assessed using support, confidence, and lift measures, and all rules obtained lift values greater than 1, indicating meaningful positive dependencies among products. The strongest rule revealed that customers who purchased Bread and Milk were highly likely to also purchase Cereal, with a confidence of 60.5% and a lift value of 2.70. Although both algorithms produced comparable rule quality, ECLAT demonstrated superior computational efficiency due to its vertical data representation approach. These findings provide practical insights for retail decision-making, particularly in cross-selling strategies, product bundling, and shelf arrangement optimization.
