The development of machine learning technology, especially in dermatology, offers excellent opportunities for classifying and diagnosing skin conditions such as acne. This study aims to apply and compare two machine learning methods, K-Nearest Neighbors (KNN) and Random Forest methods, to classify acne severity into three levels: mild, moderate, and severe. The acne density and average confidence features were extracted from facial images using the YOLOv8 model based on acne bounding boxes. While the KNN model achieves 95% accuracy, the Random Forest model reaches 97%, indicating superior performance with excellent precision, recall, and F1-score values. With its level of accuracy, the integration of the Random Forest model and the features extracted using the YOLOv8 model appear to be a promising tool in dermatology for classifying acne severity in a more accurate and effective way.
Stunting is a critical public health issue primarily caused by malnutrition, which hampers the growth of children. This study evaluates the performance of two machine learning models, K-Nearest Neighbors (KNN) and Decision Tree, in classifying stunting status in toddlers. Three strategies for handling class imbalance—no sampling, Synthetic Minority Over-sampling Technique (SMOTE), and random undersampling-are compared to enhance the detection of the minority class (stunting). The results show that KNN with SMOTE achieved the best performance, with an accuracy of 99.17% and an F1-Score of 99.17%, highlighting the model’s effectiveness in balancing sensitivity to the minority class. In contrast, although Decision Tree achieved an accuracy of 99.11% without sampling technique, it faced challenges in detecting stunting, which were addressed with the use of SMOTE, improving its accuracy to 97.41%. The application of random undersampling caused a significant decline in performance for both models. These findings underscore the effectiveness of SMOTE in handling class imbalance for stunting detection and provide valuable insights into the application of machine learning techniques in addressing public health issues.
Sign language is a vital communication method for individuals with hearing loss or deafness, with variations reflecting unique cultural contexts. Real-time recognition of sign language can bridge communication gaps, yet developing tools for Indonesian Sign Language (BISINDO) is challenging due to limited datasets. This research addresses these challenges by enhancing BISINDO detection and real-time recognition, focusing on flexible dataset collection and adaptation to varying lighting conditions. Three convolutional neural networks—InceptionV3, MobileNetV2, and ResNet50—are evaluated with optimizers SGD, Adagrad, and Adam to determine the best architecture-optimizer combination. Models were trained on a common dataset and analyzed for optimal performance. Real-time recognition uses MobileNetV2 SSD, integrating data augmentation to improve performance under diverse lighting. The system was deployed on a mobile device for practical use. Results showed the real-time model attained a mean Average Precision (mAP) of 90.34%. This study demonstrates significant advancements in BISINDO recognition and real-time application
Handling class imbalances in datasets is a significant challenge in the classification process. Disruption occurs if the minority class has a crucial role in decision-making. Oversampling is one of the solutions that is widely used to overcome this problem. This study compares the performance of three popular oversampling methods, namely SMOTE (Synthetic Minority Oversampling Technique), SMOTE-ENN (SMOTE with Edited Nearest Neighbor), and Borderline-SMOTE, based on the number of outlier data produced. Outlier data is measured using a Z-score-based statistical approach. The research was conducted by applying the three oversampling methods on several datasets. Evaluation is carried out by counting the number of outlier data after the resample process, as well as by evaluating their impact on the performance of the classification model using metrics such as accuracy, precision, recall, and F1-score. The research results show that there is no significant difference in the number of outlier data in SMOTE, ENN SMOTE, or borderline SMOTE. In the diabetes.csv dataset, it was found that the percentage of outlier data in the initial condition and the condition after resampling with SMOTE, resampling with SMOTE ENN, and borderline SMOTE were 7.4%, 6.8%, 6.7%, and 63%, respectively. For the predict_ honor.csv dataset, the data are 7.1%, 7.3%, 7.6%, and 7%. For the winequality.csv dataset, the data are 8%, 7.8%, 6.8%, and 5.8%. Meanwhile, smoking.csv data found 7.1%, 7.3%, 7.6%, and 7.0%. However, if we look at each feature in each dataset, more varied conditions are found regarding the performance of the three algorithms, which is related to the number of outlier data produced. In terms of differences, no significant differences were found in the number of outlier data produced. The second finding is related to the performance of the decision tree classification model. It can be stated that the influence of feature correlation is more important than perfect data balance in the dataset.
Strawberry plants, or Fragaria x ananassa, are shrubs in the Rosaceae (rose) family that produce sweet and scented red fruit. Strawberries are high in vitamin C and other minerals. The benefits of growing strawberries in smart greenhouses are one of the hydroponic farming sectors advances. The construction of a smart greenhouse system that can be monitored and controlled automatically simplifies agricultural research, which formerly relied on traditional farming and wet labs with long study timeframes and high expenses. This innovation makes it easier for researchers to study the impact of the Internet of Things (IoT) on strawberry plant growth by using several sensors in the greenhouse and Artificial Intelligence (AI) to save time and money in optimizing strawberry plant growth. Meanwhile, the Support Vector Machine (SVM) algorithm with a multi-classification category on leaves 3, 4, and 5 achieved Precision: 0.96, Recall: 0.95, F1-Score: 0.95, and Accuracy: 0.95. The accuracy level reaches 95%, implying that machine learning can be used in strawberry cultivation to assist hydroponic farmers.
