Green Intelligent Systems and Applications

Spam and Phishing Whatsapp Message Filtering Application Using TF - IDF and Machine Learning Methods

Ferdinand Aprillian Manurung — 2025-01-18

The rapid development of communication technology has led to an increase in the number of unwanted messages, such as spam and phishing attempts. However, this progress has not been accompanied by sufficient user awareness of the basics of technology use. Additionally, the enforcement of laws regarding internet-based crimes remains unclear, further increasing the risk for users of internet technology to fall victim to such crimes. As one of the media prone to spam and phishing, WhatsApp is the focus of this research, which aims to develop an application capable of filtering spam and phishing messages. The application employs the TF-IDF (Term Frequency-Inverse Document Frequency) method and machine learning using the Random Forest model. It is developed using the MVVM (Model-View-ViewModel) architecture, enabling the separation of business logic from the user interface, thereby improving development and maintenance efficiency. The research findings demonstrate that the combination of TF-IDF and Random Forest achieves high accuracy in classifying spam and phishing messages. Performance evaluation using a confusion matrix reveals an accuracy rate of 92%. For the safe message class, the precision, recall, and F1 scores are 89%, 95%, and 92%, respectively, while for the dangerous message class, the scores are 95%, 88%, and 92%, respectively. Furthermore, the integration of the model and application performed exceptionally well, as evidenced by black-box testing results. All test scenarios were met, successfully detecting test messages with 98% accuracy. Therefore, the developed application provides enhanced protection for WhatsApp users against digital threats.

A Benchmark Study of DeepLabV3+, U-Net++, and Attention U-Net for Blood Cell Segmentation

Clara Lavita Angelina — 2025-03-29

Cell segmentation is a critical process in biomedical image analysis. This study evaluated the performance of three state-of-the-art deep learning models—DeepLabV3+, U-Net++, and Attention U-Net—using the Blood Cell Count and Detection (BCCD) dataset, which contains annotated images of blood cells. The models were rigorously analyzed through qualitative and quantitative evaluations, employing accuracy, precision, recall, and F1 score metrics. The results demonstrated that all three models achieved high segmentation performance, with U-Net++ excelling in accuracy (0.9740), precision (0.9511), and F1 score (0.9576), Attention U-Net achieving the highest recall (0.9692), and DeepLabV3+ providing a balanced performance across all metrics. Qualitative analyses revealed that U-Net++ delivered superior segmentation of complex and overlapping cell structures, while Attention U-Net exhibited exceptional sensitivity to dense cell clusters. Training and validation curves of the models confirmed their stability and generalizability, indicating efficient convergence without overfitting. By highlighting the unique strengths of each model, this study emphasized the importance of selecting architectures tailored to specific tasks. Future research will expand the application of these models to diverse biomedical datasets to further advance automated image analysis and its impact on healthcare outcomes.

Design of A Braille Printer Based on ESP32 Microcontroller with Voice Input

Maria Beatrix — 2025-02-26

Braille is a tactile phonetic alphabet system invented by Louis Braille, a blind teacher from France, in the 1800s. The Braille system was recognized as "a vital language of communication, as valid as all other languages in the world" in 2005. There are other alternatives, such as touch-based methods, to convey information that is generally obtained through sight. One of them is the use of Braille letters for reading, writing, and improving welfare by increasing insight. However, only 52 special schools in Indonesia have printers for Braille books. Limited access to Braille printing facilities in Indonesia is due to high costs. The cost of a printer machine, approximately 50 million per school, poses a challenge in providing learning facilities. This research proposes a compact and cost-effective Braille printer using an ESP32 microcontroller with both speech and mechanical switch inputs. The mechanical switch is used for typing text to be printed, while the microphone captures sound input in the form of audio, as it is easier to use. Audio input is processed using speech-to-text technology. The speech-to-text process is carried out with speech recognition, which listens to the words spoken by the user and matches them with the data in the module to execute specific commands. This Braille printer is designed to print Braille letters based on data received directly from individuals with and without disabilities. The printer accepts input in the form of speech or text, which is then sent to the processing module, the ESP32 microcontroller. Once all data is processed, the Braille printer module controls axis movements using a stepper motor. Braille prints are embossed to create raised dots on paper. Experimental results demonstrate 100% accuracy for both speech and typing inputs, along with reliable printing performance on standard HVS paper. Compared to previous solutions, the proposed design is more versatile, affordable, and portable. This study presents a practical solution for increasing access to Braille education and information.

Fraud Classification in Online Payments Using Supervised Machine Learning Algorithms

Arda Surya Editya — 2025-03-21

Online payment systems have become a cornerstone of modern financial transactions, providing convenience and efficiency. However, the rise of such systems has also led to an increase in fraudulent activities, posing significant risks to users and service providers. This research focused on optimizing the classification of fraudulent transactions in online payment systems using supervised machine learning algorithms. This study explored the performance of several widely used algorithms, including Naïve Bayes, Decision Tree, Random Forest, Gradient Boosting Tree, and Support Vector Machine (SVM). A comprehensive dataset of online payment transactions was used to evaluate the effectiveness of these algorithms in identifying fraudulent activities. Various performance metrics, such as accuracy, precision, and F1 score, were employed to assess and compare classification capabilities. In addition, feature engineering and data preprocessing techniques were applied to improve the models’ predictive performance. The results demonstrated that, while each algorithm had its strengths, ensemble-based methods like Gradient Boosting Tree outperformed others in terms of classification accuracy and robustness. The findings highlighted the importance of selecting appropriate machine learning algorithms and fine-tuning their parameters to achieve optimal fraud detection in online payment systems. This study provides valuable insights for financial institutions and developers to enhance security measures and mitigate fraud risks in digital payment platforms.

Harnessing Smart Farming: Key Determinants of Automated Mini Greenhouse Adoption and Use in the Philippines

Eugenia R. Zhuo — 2025-02-04

This research investigated the determinants of adopting and sustaining the utilization of automated mini-greenhouses in the Philippines, a nation particularly vulnerable to climate change. Using an integrated theoretical framework combining the Unified Theory of Acceptance and Use of Technology (UTAUT2), Diffusion of Innovation (DOI), and Actor-Network Theory (ANT), this research employed a quantitative approach to assess key constructs, such as performance expectancy, effort expectancy, social influence, facilitating conditions, hedonic motivation, trust, habit, and technology readiness. Data were collected through structured surveys administered to smallholder farmers, and the results were analyzed using Python-based statistical tools. The findings indicated that performance expectancy and social influence were significant predictors of technology adoption, while habit and facilitating conditions strongly influenced continued use. Trust and resource accessibility, derived from DOI and ANT, also emerged as critical factors in sustained utilization. These results contributed to understanding smart farming adoption in the context of climate resilience and sustainable agriculture. Future research should explore broader applications of such technologies and further examine their long-term sustainability.

Comparison Of Feature Extraction Techniques For Long Short-Term Memory Models In Indonesian Automatic Speech Recognition

Dimas Dwi Armaisya — 2025-04-13

Automatic Speech Recognition (ASR) faced challenges in accuracy and noise robustness, particularly in Bahasa Indonesia. This research addressed the limitations of single feature extraction methods, such as Mel-Frequency Cepstral Coefficients (MFCC), which were sensitive to noise, and Relative Spectral Transform - Perceptual Linear Predictive (RASTA-PLP), which was less effective in frequency representation, by proposing a hybrid approach that combined both techniques using Long Short-Term Memory (LSTM) models. MFCC enhanced spectral accuracy, while RASTA-PLP improved noise robustness, resulting in a more adaptive and informative acoustic representation. The evaluation demonstrated that the hybrid method outperformed single and non-extraction approaches, achieving a Character Error Rate (CER) of 0.5245 on clean data and 0.8811 on noisy data, as well as a Word Error Rate (WER) of 0.9229 on clean data and 1.0015 on noisy data. Although the hybrid approach required longer training times and higher memory usage, it remained stable and effective in reducing transcription errors. These findings suggested that the hybrid method was an optimal solution for Indonesian speech recognition in various acoustic conditions.

Twitter Sentiment Analysis of Mental Health Issues Post COVID-19

Panca Dewi Pamungkasari — 2025-03-29

The Coronavirus Disease 2019 (COVID-19) impacted many aspects of daily life, including mental health, as some individuals struggled to adjust to the rapid changes brought on by the pandemic. This paper investigated sentiment analysis of Twitter data following the COVID-19 pandemic. Specifically, we analyzed a large corpus of tweets to understand public sentiment and its implications for mental health in the post-pandemic context. The Naïve Bayes and Support Vector Machine (SVM) classifiers were used to categorize tweets into positive, negative, and neutral sentiments. The collected tweet data samples showed that 38.35% were neutral, 32.56% were positive, and 29.09% were negative. Results using the SVM method showed an accuracy of 84%, while Naïve Bayes achieved 80% accuracy.