Publication Search

The growth of e-commerce in Indonesia has led to the emergence of various online shopping platforms, with Shopee being one of the most popular in Semarang City. User reviews on the Shopee application serve as a valuable data source for analyzing customer satisfaction levels; however, the large volume of data requires a systematic and accurate analytical approach. This study aims to analyze user review sentiments of the Shopee application using three machine learning algorithms: Random Forest, Naïve Bayes, and Support Vector Machine (SVM), as well as comparing the accuracy of these three algorithms. This research utilized 1000 reviews collected through web scraping from the Play Store, which were categorized into three classifications: positive, neutral, and negative sentiments. The analysis process encompassed pre-processing stages, feature extraction using TF-IDF, and classification using Random Forest, Naïve Bayes, and Support Vector Machine algorithms. The results demonstrated that the Random Forest algorithm achieved the highest accuracy at 96.19%, followed by Support Vector Machine with 95.71% accuracy, and Naïve Bayes with 84.76% accuracy. This research highlights the effectiveness of Random Forest and SVM in classifying user review sentiments towards the Shopee application.

Computer vision technology has advanced rapidly and made significant contributions across various fields, including object identification in images. This study aims to develop a computer vision-based system to identify fruit types from images. A machine learning model is applied using a dataset of fruit images to train the system for accurate fruit recognition. The primary processes include data acquisition, image preprocessing, feature extraction, model training, and performance evaluation. The results demonstrate a high level of accuracy in identifying specific fruit types, showcasing the potential of this technology in agricultural and commercial applications.

This study presents an advanced approach to multi-class skin lesion classification by leveraging an ensemble model comprising the Inception-V3, ResNet-50, and VGG16 architectures. The classification task focuses on categorizing skin lesions into distinct classes, including Melanoma, basal cell carcinoma (BCC), and squamous cell carcinoma (SCC), using the ISIC dataset, a comprehensive collection of dermoscopic images. In order to properly balance the dataset, the oversampling strategy is utilized, as some lesion types are underrepresented due to inherent imbalances in the dataset. By ensuring that the model is trained on a more representative dataset, this balancing improves the algorithm's capacity to categorize all lesion types properly and impartially. By combining the complementary features of ResNet-50, Inception-V3, and VGG16, the ensemble technique improves the overall classification performance. ResNet-50 is chosen for its deep feature extraction capabilities, which help capture fine details in lesion patterns. Inception-V3 is selected for its multi-scale processing, allowing it to effectively analyze lesions at varying resolutions and sizes. VGG16 is included due to its simple yet highly effective architecture for image classification tasks. The ensemble model with data augmentation significantly outperforms individual models in skin lesion classification for both the original and balanced ISIC datasets regarding accuracy, precision, recall, and F1-score. This method offers a robust solution for skin lesion classification, contributing to more accurate and reliable diagnostic tools in dermatology.

Abstract. This study presents a web-based system for identifying traditional herbal leaves using K-Nearest Neighbors (KNN) and image processing techniques focused on analyzing leaf shape and color. The dataset used consists of images of various types of herbal leaves, providing a basis for classification and medicinal benefit information retrieval. The system was tested with multiple leaf samples to assess accuracy, speed, and effectiveness in identifying leaf types based on visual characteristics. Results show that the system can recognize different types of herbal leaves and display information on their medicinal properties in a user-friendly interface..

This study presents the implementation of Transfer learning using the ResNet-18 architecture for classifying 10 musical instrument categories based on visual representations of audio signals. The audio waveform is transformed into image-like inputs appropriate for CNN processing, accompanied by data augmentation and ImageNet-standard normalization. ResNet-18 is utilized due to its efficient feature extraction capability enabled by residual blocks, which help overcome vanishing gradient issues. The model was trained for 10 Epochs using the AdamW optimizer and Cross-Entropy Loss. Experimental results show that the model achieved a maximum validation accuracy of 77.35%, with a stable downward trend in training loss, indicating effective feature learning. However, several misclassification cases were observed, particularly among instruments with similar spectral characteristics, such as drum–violin and tabla–sitar. These findings demonstrate that while ResNet-18 performs reliably for musical instrument classification, further improvements remain possible through deeper architectures like ResNet-50, more comprehensive hyperparameter optimization, and the use of richer audio representations such as Mel-Spectrograms. This research provides an essential foundation for developing automated music analysis systems powered by Deep Learning.

Image classification is widely used in everyday life such as in car steering, closed-circuit television (CCTV), traffic cameras, etc. The implementation of image classification can be done using several methods, including neural network and support vector machine models. The neural network method is able to find the right weights that allow the network to show the desired behaviour while the support vector machine method has many dimensions and can overcome linear and non-linear data. In this research, feature extraction was carried out using VGG16 to increase accuracy. This research aims to find out how to implement the neural network and SVM algorithms to classify images and determine the results of analyzing the performance of the two methods. The data used in this study is secondary data consisting of 10 types of large wild cats with a total of 2339 training image datasets and 50 testing image datasets. The research stages consist of data augmentation, model design, model training, and model evaluation. Classification with the neural network model produced an accuracy of 96% and the support vector machine model produced an accuracy of 96%, which means that in a consistent training environment, the two models have the same performance.

Handling of dental disease problems requires that it be handled quickly and correctly, but not all teams of dental experts can carry out treatment quickly due to the lack of a team of dental experts who are in the workplace or hospital 24 hours a day.  Apart from that, the public also has very little knowledge of information about dental disease, so that to treat dental disease, people have to consult a dentist. To classify images of dental disease, feature extraction is needed. Feature extraction is taking characteristics of an object that can describe the image. One example of image feature extraction used is Red, Green, Blue (RGB). This feature extraction is often used to identify or classify an image. Dental image data that will be used in the classification process are tooth abrasion, anterior crosbite, cavities and gingivitis. K-Nears Neigbor is the simplest data mining algorithm.  The aim of this algorithm is to find the results of the closest distance classification for each object.  In determining the distance, the data is initially divided into two parts, namely training data and testing data. After receiving the training data and testing data, the distance from each testing data (Equilidence Distance) to the training data is calculated. The K-Nearest Neighbors method can be applied to classify dental disease based on images of types of dental disease using Matlab software. As a result of the image data training process, 40 image data were input, training results obtained were 100%.

The Internet of Things (IoT) is an innovative technology that makes our environment smarter, with IoT devices as an integral part of home automation. Smart home systems are becoming increasingly popular as an IoT service in the home that connects via a network. Due to the security weakness of many devices, the malware is targeting IoT devices. After being infected with malicious attacks on smart devices, they act like bots that the intruders can control. Machine learning methods can assist in improving the attack detection process for these devices. However, the irrelevant features raise the computation time as well as affect the detection accuracy in the processing with many features. We proposed a machine learning-based IoT security framework using feature correlation. The feature extraction scheme, one-hot feature encoding, correlation feature selection, and attack detection implement an active detection mechanism. The results show that the implemented framework is not only for effective detection but also for lightweight performance. The proposed system outperforms the results with the selected features, which have almost 100% detection accuracy. It is also approved that the proposed system using CART is more suitable in terms of processing time and detection accuracy.

Fingerprint recognition is a popular biometric technology due to its unique properties and high accuracy rate. Fingerprint recognition systems generally use fingerprint image representations, such as grayscale images, phase images, skeleton images, and minutiae. In this research, fingerprint image pre-processing is performed using Gaussian Blur, Median Blur, Thresholding, Otsu Thresholding, Thinning with Guo-Hall algorithm, and Minutiae Detection. Minutiae detection produces 426 termination points and 459 bifurcation points. The results of the pre-processing and minutiae detection were then used for minutiae matching on 5 different images. Minutiae matching produces varying degrees of similarity with a high level of accuracy, reaching an average accuracy of 88.80%.

The advancement of e-commerce has changed the way people shop. However, there is a mismatch between the actual quality of a product and the seller’s description. Product reviews are an important source of information for making purchasing decisions. However, processing large numbers of reviews manually is difficult. This research aims to detect emotions in Indonesian language product review texts using contextual embeddings. The public dataset used was PRDECT-ID, which comprises five emotion labels. The methods used include data preprocessing, feature extraction using contextual embeddings such as Bidirectional Encoder Representations from Transformers (BERT), and classification using Decision Tree, Naïve Bayes, and k-Nearest Neighbors (KNN). Among the compared models, the KNN model demonstrated the highest improvement, achieving a 15.09% enhancement over the decision tree results. This research provides insights into the effectiveness of contextual embeddings in detecting emotions in Indonesian language product review texts.

This research aims to classify news headlines into clickbait and non-clickbait using the Multinomial Naive Bayes method. The data used comes from the dataset CLICK-ID: A Novel Dataset for Indonesian Clickbait Headlines. The research process involves stages of data collection, preprocessing, feature extraction, model training, model evaluation, and result analysis. The test results show that the Multinomial Naive Bayes algorithm consistently produces an accuracy rate of around 78%. Optimization using Grid Search did not result in an accuracy improvement. However, there was an improvement in the recall value for the non-clickbait class from 76% to 80%. The best parameter found was an alpha of 0.15. Therefore, the Multinomial Naive Bayes algorithm can be effectively used to address the problem of classifying clickbait news headlines, with the potential to contribute to clickbait prevention efforts in the future.

The process of identifying the ripeness level of cayenne peppers is an important step in cultivation and post-harvest handling. Dependence on the quality factors of farmers, such as visual diversity and differences in ripeness perception, results in subjective harvest outcomes. This manual process is also prone to inconsistent results, as humans have time limitations, fatigue, and sometimes lack concentration when sorting for long periods. To minimize these issues, technological intervention is needed to mechanically classify the ripeness level of cayenne peppers. This research aims to develop a classification model for the maturity level of cayenne pepper plants. This research proposes the use of the CNN method for feature extraction and KNN for data classification based on the features extracted by CNN. From the test scenarios carried out, the classification carried out by KNN based on CNN feature extraction got the best accuracy of 99.33%, while the CNN classification model got the best accuracy of 87.33%.

This research designs a system to classify apple leaf diseases using RGB (red, green and blue) color feature extraction. The essence of this research is to design a system to recognize and determine disease on apple leaves based on RGB color features using the Matlab 2024 application. The data in this research uses apple leaf images from kaggle.com, which are then cropped and adjusted to the image shape and precision in the leaf image. , Increasing the contrast of the cropped image and converting it to a grayscale image, Determining the threshold for binarization and converting the grayscale image to a binary image, Detection of green, yellow, and black/gray pixels based on RGB values ​​and calculating the proportion of each color, Detection of pixels scab by filtering out black/grey pixels that do not include green or yellow pixels Classification of leaves based on the proportion of detected colors. With the method that has been passed and uses apple leaf data, namely Healthy, Rust and Scab, each data contains 20 images with a total of 60 images and the level of accuracy is determined using the labeling method for each data and reaches the final result with an accuracy level of 86, 6667% which has a fairly accurate meaning  

This research explores the classification of brass instrument sounds using Convolutional Neural Network (CNN) combined with Mel-Frequency Cepstrum Coefficient (MFCC) feature extraction. This research aims to improve the accuracy of brass instrument sound recognition by utilizing CNN's ability to process audio data. Through experiments conducted with different audio durations and variations in CNN model architecture, this study evaluates the impact of dataset separation and model design on classification performance. The results show that dataset duration and CNN model architecture significantly affect classification accuracy, with the highest accuracy achieved in the scenario using 30 seconds of audio duration with an accuracy value of 84%. In addition, experiments varying the number of convolution layers in the CNN model show that the selection of the model architecture plays an important role in classification performance. Overall, this research contributes to advancing the field of audio classification by providing insight into the optimal dataset duration and model architecture for wind instrument speech recognition using CNNs.

This study proposes an approach for human fall classification utilizing a combination of Weighted Moving Average (WMA) and Convolutional Neural Networks (CNN) on the SisFall dataset. Falls among elderly individuals pose a significant public health concern, necessitating effective automated detection systems for timely intervention and assistance. The SisFall dataset, comprising accelerometer data collected during simulated falls and activities of daily living, serves as the basis for training and evaluating the proposed classification system. The proposed method begins by preprocessing accelerometer data using a WMA technique to enhance signal quality and reduce noise. Subsequently, the preprocessed data are fed into a CNN architecture optimized for feature extraction and fall classification. The CNN leverages its ability to automatically learn discriminative features from raw sensor data, enabling robust and accurate classification of fall and non-fall events. Experimental results demonstrate the efficacy of the proposed approach in accurately distinguishing between fall and non-fall activities, achieving high classification performance metrics such as accuracy, precision, recall, and F1-score. Comparative analysis with existing methods showcases the WMA-CNN hybrid approach's superiority in classification accuracy and robustness. Overall, the proposed methodology presents a promising framework for real-time human fall classification using sensor data, offering potential applications in wearable devices, ambient assisted living systems, and healthcare monitoring technologies to enhance safety and well-being among elderly individuals.

Sabu woven fabric is one of the cultural heritages of Sabu Island. In addition to being a cultural heritage, Sabu woven fabric is one of the handicrafts that still exist today which is preserved by Sabu women. Based on its manufacture, the classification process of Sabu woven fabric is based on color or motif identification. However, the classification process is not an easy process, because the classification process requires time and experts in the field of Sabu woven fabric. In addition to the classification process, the wider community also does not get much information about Sabu woven fabric clearly, because it is necessary to introduce the type of Sabu woven fabric, so that people can know or recognize the type of Sabu ikat woven fabric based on its type. Digital image processing techniques are utilized to build a system that can overcome the problems faced. Furthermore, image feature extraction will be carried out using gray level co-occurrence matrix (GLCM) with 4 features namely contrast, correlation, energy, and homogeneity with angles of 0°, 45°, 90°, and 135°. Each GLCM feature shows the same value even though the original image is rotated. After image feature extraction, the extracted data will be classified using the TensorFlow library. From these results it can be concluded that the program succeeded in selecting the type of Sabu ikat woven fabric class.

The rapid advancement of deep learning technologies has significantly transformed healthcare analytics, particularly in medical data prediction and classification. This study proposes a hybrid Convolutional Neural Network–Long Short-Term Memory (CNN–LSTM) framework for multi-modal healthcare data analysis, integrating medical imaging, structured electronic health records (EHRs), and IoT-generated time-series physiological signals. The proposed architecture combines spatial feature extraction through CNN with temporal dependency modeling via LSTM to enhance predictive accuracy and clinical decision support. A quantitative experimental design was employed, utilizing multi-source healthcare datasets that underwent preprocessing, normalization, and feature engineering prior to model training. The performance of the hybrid model was evaluated using Accuracy, Precision, Recall, F1-Score, AUC-ROC, and Mean Absolute Error (MAE), and compared with conventional machine learning models and standalone deep learning architectures. Experimental results demonstrate that the proposed CNN–LSTM model achieves superior performance, with improved classification accuracy and reduced prediction error, while maintaining strong generalization capability. The findings indicate that integrating spatial and temporal feature learning significantly enhances disease detection, risk stratification, and personalized treatment planning. This approach supports the development of intelligent clinical decision support systems and scalable smart healthcare environments. The proposed framework offers a reliable and efficient solution for advanced healthcare analytics in IoT-enabled systems.

In the evolving landscape of agricultural technology, recognizing rice diseases through computational models is a critical challenge, predominantly addressed through Convolutional Neural Networks (CNN). However, the localized feature extraction of CNNs often falls short in complex scenarios, necessitating a shift towards models capable of global contextual understanding. Enter the Vision Transformer (ViT), a paradigm-shifting deep learning model that leverages a self-attention mechanism to transcend the limitations of CNNs by capturing image features in a comprehensive global context. This research embarks on an ambitious journey to refine and adapt the ViT Base(B) transfer learning model for the nuanced task of rice disease recognition. Through meticulous reconfiguration, layer augmentation, and hyperparameter tuning, the study tests the model's prowess across both balanced and imbalanced datasets, revealing its remarkable ability to outperform traditional CNN models, including VGG, MobileNet, and EfficientNet. The proposed ViT model not only achieved superior recall (0.9792), precision (0.9815), specificity (0.9938), f1-score (0.9791), and accuracy (0.9792) on challenging datasets but also established a new benchmark in rice disease recognition, underscoring its potential as a transformative tool in the agricultural domain. This work not only showcases the ViT model's superior performance and stability across diverse tasks and datasets but also illuminates its potential to revolutionize rice disease recognition, setting the stage for future explorations in agricultural AI applications.

Disease control of chili leaf citra plants is an important aspect in modern agriculture to increase crop yields and reduce losses due to pest attacks on chili leaf citra plants. In this research, identification of chili leaf diseases uses Gray Level Co-Occurrence to obtain image features, and the Support Vector Machine (SVM) method is used to classify the feature extraction results according to leaf disease categories in the test image. Based on the disease class using the test image. .As a classification tool for identifying plant pests in images of chili leaves, the dataset used in this research consists of images of leaves that represent normal conditions and conditions attacked by pests. The pest identification process consists of several stages, including image pre-processing, feature extraction, as well as training and testing. SVM model.