HHHerlawati Herlawati

Herlawati Herlawati

*This is an algorithmically generated author record based on publicly available information.

Herlawati Herlawati has a total of 116 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Komparasi Kinerja ResNet50 dan MobileNetV2 pada Klasifikasi Citra Multi-Domain

    by Herlawati Herlawati, Andy Achmad Hendharsetiawan, Wowon Priatna, Afina Putri Dzulqiyana, Regita Ari Rahmadanti

    This study aims to compare the performance of the Convolutional Neural Network (CNN) architectures ResNet50 and MobileNetV2 in multi-domain image classification, namely herbal leaf images and facial skin undertone images. The herbal plant dataset consists of 11 classes of leaf images, while the skin undertone dataset includes warm, cool, and neutral categories obtained from secondary and primary data sources. The research stages include image preprocessing, model training using transfer learning, and performance evaluation using accuracy, precision, recall, F1-score, and confusion matrix. The results show that ResNet50 achieved better performance than MobileNetV2 in both classification domains. In herbal plant classification, ResNet50 achieved an accuracy of 95.00%, while MobileNetV2 obtained 94.09%. In facial skin undertone classification, ResNet50 achieved an accuracy of 89.3% and a weighted F1-score of 0.893, whereas MobileNetV2 achieved an accuracy of 68.3% and a weighted F1-score of 0.684. These findings indicate that ResNet50 is more effective and stable than MobileNetV2 for multi-domain image classification.

  • Evaluasi Kinerja Metode Content-Based Filtering pada Sistem Rekomendasi Destinasi Wisata dan Rute Pendakian Berbasis Web

    by Herlawati Herlawati, Andy Achmad Hendharsetiawan, Bertnardo Mario Uskono, Acah Acah

    This study aims to evaluate the Content-Based Filtering method in a web-based recommendation system for tourist destinations and hiking routes. The study integrates two domains: 1,621 tourist destinations in Bogor and 126 mountain hiking routes across Java. Tourism data were collected through web scraping from Google Maps, while hiking route data were obtained from Gunungbagging.com and Muncak.id. The attributes used include category, location, ticket price, facilities, rating, distance, elevation gain, estimated duration, difficulty level, and recommended user experience. Hiking duration was estimated using the Naismith Rule. The system constructs a user profile based on user preferences and computes the similarity between the user profile and item characteristics using Cosine Similarity. For tourism data, a Deep Autoencoder was employed to reduce 54 features into 8 latent dimensions, while the Haversine Formula was applied as a geospatial filter to account for location proximity. Evaluation using Precision@10 yielded average scores of 0.80 for hiking route recommendations and 0.85 for tourist destination recommendations. The results indicate that Content-Based Filtering effectively produces relevant and precise recommendations that align with user preferences and supports more objective and efficient decision-making.

  • Perbandingan Kinerja Random Forest dan Long Short-Term Memory dalam Prediksi Kedalaman Hiposentrum Gempa di Maluku dan Laut Banda

    by Laila Salsabilla Hanifa, Herlawati Herlawati, Rakhmat Purnomo

    Hypocenter depth is an important parameter in earthquake analysis because it is related to the earthquake source mechanism and its impact on the Earth’s surface. The Maluku and Banda Sea regions have high seismic activity caused by interactions among active tectonic plates. This study compares the performance of Random Forest and Long Short-Term Memory (LSTM) algorithms in predicting earthquake hypocenter depth using historical earthquake data from the United States Geological Survey (USGS). Model evaluation used MAE, RMSE, and R² metrics. The results show that Random Forest outperformed LSTM, achieving MAE, RMSE, and R² values of 31.94, 62.83, and 0.73, respectively. Meanwhile, LSTM produced MAE of 69.35, RMSE of 119.31, and R² of -0.11. Therefore, Random Forest was considered more effective for predicting earthquake hypocenter depth in the Maluku and Banda Sea regions.

  • Evaluasi Kinerja Clustering K-Means Menggunakan Elbow Method dan Silhouette Coefficient pada Data Penimbangan Sampah

    by Anis Athifah, Herlawati Herlawati, Fata Nidaul Khasanah

    Bank Sampah Perumahan Sapta Pesona RW 08 Bekasi City has waste weighing data that is still limited to administrative records and has not been optimally used to identify waste characteristic patterns. This study aims to Cluster waste data based on quantity/tonnage and unit price using the K-Means algorithm. The research method used Cross Industry Standard Process for Data Mining (CRISP-DM). The dataset consisted of 810 waste weighing records from January 2024 to December 2025 with item type, quantity/tonnage, and unit price attributes. The number of Clusters was determined using the Elbow Method based on the Sum of Squared Error (SSE), while the Clustering results were evaluated using the Silhouette Coefficient. The results show that the data were grouped into three Clusters with a Silhouette Coefficient value of 0.7844, indicating a strong Cluster structure. Cluster 0 represents waste with low to moderate tonnage and unit price, Cluster 1 represents waste with low tonnage and high unit price, while Cluster 2 represents waste with high tonnage and low to moderate unit price. These results help managers understand waste data characteristics more systematically

  • Open Access

    Penyuluhan Dan Pelatihan Keamanan Data Digital Bagi Masyarakat Dari Ancaman Kejahatan Cyber

    by Herlawati Herlawati, Rafly Fandiansyah, Mirza Cahya Ningrum, Caroline Julyana Magdalena, Muhammad Gymnastiar, Naufal Eka Wicaksono, Naufal Arif Fadilah, Raka Rismayana, Muhammad Reinaldy Santoso, Putra Aldi Purnama, et al.

    The rapid development of information technology has led to increased use of digital media by the community, which has not been accompanied by adequate understanding of digital data security. This condition makes the community vulnerable to various cybercrime threats such as online fraud, personal data theft, phishing, and the spread of hoaxes. This community service activity aims to improve the digital literacy and awareness of residents of Taman Kebayoran Housing Area RW 013, Setiamekar Village, Bekasi Regency, regarding the importance of digital data security through counseling and training programs. The implementation methods include initial socialization, face to face counseling, digital data security training, and the development of a Housing Profile Website as an official and verified information medium. The results show an increase in community understanding of types of cyber threats, methods to identify false information, and preventive measures against digital fraud. In addition, the existence of the housing profile website is considered effective in supporting information transparency and strengthening communication between community administrators and residents. The originality of this activity lies in the integration of community based cybersecurity education with the development of a local digital platform that functions as a literacy and official information medium. This program is expected to serve as a sustainable community service model in building a safe and responsible digital ecosystem at the neighborhood level.

  • Analisis Sentimen Ulasan Produk Sneakers Lokal Pada Tokopedia Menggunakan Algoritma Naïve Bayes dan Support Vector Machine

    by I Komang Arya Trisumeikra, Herlawati Herlawati, Agus Hidayat

    Transformation in various sectors, including the fashion industry, especially sneakers. Sneakers are now a symbol of modern lifestyle and global trends, with brands such as Nike and Adidas dominating the market. However, high prices are an obstacle for many Indonesian consumers. This opens up opportunities for local brands to offer quality products at affordable prices through e-commerce such as Tokopedia, the second highest traffic platform in Indonesia. The research analyzed sentiment from 1,032 consumer reviews of local sneakers from five stores: NAH Project, Aerostreet, Geoff Max, Ventela, and Brodo. The analysis was conducted using Naïve Bayes and Support Vector Machine (SVM) algorithms. The SVM evaluation results produced the highest accuracy of 98%, compared to Naïve Bayes which reached 96%. This best model is implemented in a web-based application to analyze the sentiment of new reviews, to assess the perceived quality and consumer satisfaction of local sneakers products on Tokopedia.

  • Model Prediksi Kondisi Kesehatan dari Data Medical Check-Up Menggunakan K-Nearest Neighbors dan Decision Tree

    by Tata Arya Cahyaaty, Herlawati Herlawati, Andy Achmad Hendhar Setiawan

    Medical Check-Up (MCU) is an essential procedure for the early detection of health disorders. However, manual analysis of MCU results requires time and may be subject to the interpretation of medical personnel. This study aims to develop an automatic classification system to predict health conditions based on MCU results using the K-Nearest Neighbors (KNN) and Decision Tree algorithms. The MCU data used includes blood pressure, body temperature, heart rate, as well as heart and blood pressure assessments. The models were trained and evaluated using the CRISP-DM methodology. The results show that the Decision Tree achieved an accuracy of 91.31%, while KNN achieved an accuracy of 89.75%. This system is implemented as a web-based application with a simple user interface to support the early diagnosis process at RS EMC Cibitung.

  • Open Access

    Segmentasi Berbasis Data Time Series Penjualan Produk Kopi Menggunakan Algoritma K-Means

    by Meri Anggaini, Herlawati Herlawati, Rakhmat Purnomo

    Coffee shops are businesses in the Food and Beverage (F&B) sector that contribute 7.15% to Indonesia's economy. The high demand for coffee has led to increasingly fierce competition. Kanae Coffee & Space in Bekasi faces challenges in maintaining customer loyalty and managing unpredictable demand. This study aims to apply the K-Means algorithm to cluster coffee products based on time series sales data, using the 6-step CRISP-DM methodology. The number of clusters was determined using the elbow method and confirmed with a silhouette coefficient of 0.5916 (good structure). The analysis resulted in five clusters with distinct characteristics: Cluster 0 (very low demand, stable trend, very high price), Cluster 1 (very high demand but sharply declining trend, very low price), Cluster 2 (moderately high demand, moderately stable trend, moderate price), Cluster 3 (moderate demand, slowly declining trend, moderately high price), and Cluster 4 (low demand, stable trend, moderately low price). These segmentation results are expected to serve as the basis for more effective marketing strategies and product management.

  • Open Access

    Prediksi Curah Hujan Di Kabupaten Bogor Menggunakan Long Short-Term Memory Dan Gemma 2

    by Indra Wijaya, Herlawati Herlawati, Rafika Sari

    Bogor Regency is an area that often experiences prolonged rainfall, especially during the rainy season. High rainfall causes problems such as floods and landslides. Therefore, accurate rainfall prediction is important for various needs, especially in disaster mitigation. This study aims to implement the Long Short-Term Memory (LSTM) algorithm as a model for prediction of historical rainfall data and use the Large Language Model (LLM) GEMMA 2 to provide interpretation of prediction results and recommendations based on the prediction results. The methods used include data collection from the BMKG online data website totaling 1804 data, data pre-processing, model building, model performance evaluation, and interpretation of results using LLM. The results of this study show that LSTM is able to produce the best performance by showing MSE 201.92 mm², Root Mean Square Error (RMSE) of 14.21 mm. the RMSE value shows an average error of 14.21 mm. In addition, the interpretation provided by LLM GEEMA 2 to help understand the prediction and provide practical recommendations for disaster mitigation due to rainfall.

  • Open Access

    Pendeteksian dan Klasifikasi Sampah pada Bank Sampah Berbasis Web Menggunakan YOLOv11

    by Marsyanda Salsa Nabila, Herlawati Herlawati, Agus Hidayat

    The problem of poorly managed household waste management can increase the burden on the environment and reduce the effectiveness of recycling. Waste banks in general still rely on manual systems in sorting waste which is prone to errors and requires more labor. This research aims to develop a web-based waste detection and classification system using the You Only Look Once (YOLO) version 11 yolov11n (nano) method. The research method included downloading the main secondary dataset named R1 Test version 15 from the Roboflow Universe platform, collecting other secondary datasets from internet scraping and manual photography, which resulted in a total of 27,400 images of trash with nine different types, namely bottle, cans, cardboard, cup, foil, food, paper, paper_bag, and plastic.The results show that the yolov11n model is able to detect objects with sufficient accuracy and light computational resources by producing a precision value of 91,7%, recall of 89%, mAP50 of 93,2% and mAP50-95 of 75,8% in all classes. The best model results obtained are integrated into the web using the flask framework.