Journal of Applied Informatics and Computing

DDoS Attacks Detection With Deep Learning Approach Using Convolutional Neural Network

Rafiq Amalul Widodo — 2024-08-13

The detection system of DDoS (Distributed Denial-of-Service) attacks aims to enhance network security across all facets of internet technology utilization. One is at SPKLU, which stands for Public Electric Vehicle Charging Station. The research employed a deep learning approach utilizing a Convolutional Neural Network (CNN) on a publicly available dataset. Based on our study and analysis, CNN has a precision rate of 95%. Its high accuracy and balanced performance across diverse attack types indicate the model's practical application in real-life situations. The model demonstrates promising performance in detecting different network traffic anomalies, offering significant insight into its potential for practical use. Further investigation is necessary to strengthen the resilience of DDoS assault tactics against emerging dangers and to tackle any potential constraints.

Classification of Brain Tumors by Using a Hybrid CNN-SVM Model

Talitha Safa Nabila — 2024-08-13

Brain tumors are diseases that involve the growth of brain cells, causing abnormalities in the brain region. An MRI scan is a useful tool for tumor detection. Researchers can process the obtained image data to conduct research capable of detecting brain tumor disease. Classifying brain tumors facilitates effort, planning, and accurate diagnosis, enabling the formulation and evaluation of treatment options for a patient with a brain tumor. The research was conducted to classify whether or not there was a tumor in the brain by using a combination of algorithms, namely CNN, to extract features from image data and then use SVM as a classification. CNN is a popular algorithm that deals very effectively with the complexity and variation of image data, whereas SVM is an algorithm for classification that maximizes margins and generalizations to produce accurate classifications. The project's goal is to create a hybrid model that can classify two labels based on image preprocessing processes, feature extraction, and brain tumor image data classification. In this study, the results of the CNN-SVM hybrid were able to obtain the highest score with Adam optimization and learning rate 0.001, accuracy of 98.92%, precision 98.92%, recall 98.92%, and f1-score 98.92%.

Bagging Nearest Neighbor and its Enhancement for Machinery Predictive Maintenance

Muhammad Irfan Arisani — 2024-08-13

K-nearest Neighbor is a simple algorithm in Machine learning for such a prediction classification task which plays in valuable aspects of understanding big data. However, this algorithm sometimes does a lacking job of classification tasks for many different dataset characteristics. Therefore, this study will adopt enhancement methods to create a better performance of the nearest-neighbor model. Thus, this study focused on nearest neighbor enhancement to do a binary classification task from the extremely unbalanced dataset of a machine failure problem. Firstly, this study will create new features from the machinery dataset through the feature engineering processes and transform the chosen numerical features with standardization steps as the proper scaling. Then, the modified under-sampling method will be given which will reduce the amount of the majority class to 4.75 times that of the minority class. Next is the applied grid-search tuning which will find the right parameter combinations for the nearest-neighbor model being applied. Furthermore, the previous pre-processing steps will be combined with an additional bagging method. Finally, the resulting bagged KNN will present a 0.971 rate of accuracy, 0.555 rate of precision, 0.781 rate of recall, 0.649 rate of f1-score, 0.95 auc of ROC curve, and 0.702 auc of precision-recall curve.

Facial Expression Recognition using Convolutional Neural Networks with Transfer Learning Resnet-50

Annisa Ayu Istiqomah — 2024-08-29

Facial expression recognition is important for many applications, including sentiment analysis, human-computer interaction, and interactive systems in areas such as security, healthcare, and entertainment. However, this task is fraught with challenges, mainly due to large differences in lighting conditions, viewing angles, and differences in individual eye structures. These factors can drastically affect the appearance of facial expressions, making it difficult for traditional recognition systems to consistently and accurately identify emotions. Variations in lighting can alter the visibility of facial features, while different angles can obscure critical details necessary for accurate expression detection. This study addresses these issues by employing transfer learning with ResNet-50 and effective pre-processing techniques. The dataset consists of grayscale images with a 48 x 48 pixels resolution. It includes a total of 680 samples categorized into seven classes: anger, contempt, disgust, fear, happy, sadness, and surprise. The dataset was divided so that 80% was allocated for training and 20% for testing to ensure robust model evaluation. The results demonstrate that the model utilizing transfer learning achieved an exceptional performance level, with accuracy at 99.49%, precision at 99.49%, recall at 99.71%, and an F1-score of 99.60%, significantly outperforming the model without transfer learning. Future research will focus on implementing real-time facial recognition systems and exploring other advanced transfer learning models to further enhance accuracy and operational efficiency.

Prediction of Basic Commodity Prices at the Cooperative, SME, and Trade Office Using the Least Squares Method

Desy Purwani — 2024-08-29

The Cooperatives, Small and Medium Enterprises (SME), and Trade Office is responsible for managing national affairs related to maintaining the prices of basic commodities, including the implementation of technical instructions and regulations. A significant issue faced by the office is the lack of accessible information on estimated prices of basic commodities for both the public and government. This gap primarily stems from the absence of an information system in the Pematang Siantar City area capable of publishing these estimates. The purpose of this study is to design and develop a web-based system for predicting basic commodity prices, which will record annual price fluctuations for various basic commodities at the Cooperative, SME, and Trade Office. The findings of this study will provide policymakers with a better understanding of commodity prices in traditional markets within Pematang Siantar City, serving as a foundation for future price estimations. This is particularly relevant for market operations aimed at controlling unreasonable price increases. The Least Squares method was employed to calculate the estimated prices, with the system achieving a Mean Absolute Percentage Error (MAPE) of 14.20%, indicating that the system can predict market prices with a reasonable degree of accuracy.

Improving Panic Disorder Classification Using SMOTE and Random Forest

Dini Nurmalasari — 2024-10-04

Panic disorder is a serious anxiety disorder that can significantly impact an individual's mental health. If left undetected, this disorder can disrupt daily life, social relationships, and overall quality of life. Early detection and intervention are crucial for managing panic disorder and improving the well-being of those affected. Technology plays a pivotal role in facilitating early detection through data-driven approaches that employ algorithms to identify patterns of behavior or symptoms associated with panic disorder. Accurate classification of panic disorder is crucial for effective diagnosis and treatment. However, machine learning models trained on imbalanced datasets, such as those containing panic disorder patients, are prone to overfitting, leading to poor generalization performance. This study investigates the effectiveness of the Synthetic Minority Oversampling Technique (SMOTE) in addressing overfitting in panic disorder dataset classification using the Random Forest algorithm. The results demonstrate that SMOTE significantly improves the classification performance of Random Forest. By mitigating overfitting and improving generalization to unseen data, SMOTE increases accuracy by 15 percentage points. Before using SMOTE, the accuracy was 82%, and after using SMOTE it is 97%. The findings underscore the promise of SMOTE as a tool for boosting the performance of machine learning algorithms in classifying panic disorder from imbalanced data.

Predicting Startup Success Using Machine Learning Approach

Icha Wahyu Kusuma Ningrum — 2024-10-04

Predicting startup success is important because it helps investors, entrepreneurs, and stakeholders allocate resources more efficiently, minimize risks, and enhance decision-making in an uncertain and competitive environment. Therefore, investors need to predict whether a startup will succeed or fail. Investors conduct this assessment to determine if a startup is worthy of funding. The company's founders mark success here by receiving a sum of money through the Initial Public Offering (IPO) or Merger and Acquisition (M&A) process. If the startup closes, we will consider it a failure. The data used consists of 923 startup companies in the United States. We carried out the classification using four methods: Random Forest, Support Vector Machines (SVM), Gradient Boosting, and K-Nearest Neighbor (KNN). We then compare the results from the four methods with and without feature selection. We determine the feature selection based on the relative importance of each method. The results of this study indicate that the Random Forest method with feature selection has the best accuracy, precision, recall, and F1 score than the other methods, respectively 81.85%, 80.19%, 87.09%, and 83.44%.

Sentiment Analysis of Social Media X in the 2024 Indonesian Presidential Election Using the Naive Bayes Algorithm: Candidates' Backgrounds and Political Promises

Santi Prayudani — 2024-10-04

In 2024, Indonesia holds a presidential election, and the candidates are making promises to each other to attract voters. Many people gave their opinions on X. This study uses the Naïve Bayes algorithm to analyze the sentiment of these tweets, with the aim of understanding the background of the candidates and their campaign promises. Data is collected from X by crawling technique, then data is pre-processed, trained using Naïve Bayes model, and evaluated for accuracy. Sentiments in tweets were classified as positive, negative, or neutral. The results showed that the Prabowo Subianto - Gibran Rakabuming Raka pair was the most talked about with 1005 tweets, followed by Anis Rasyid Baswedan - Muhaimin Iskandar with 707 tweets, and Ganjar Pranowo - Mohammad Mahfud M.D. with 572 tweets. The Prabowo Subianto - Gibran Rakabuming Raka pair received the most positive sentiment, which was 446 more than the other candidates.

Interpretable Machine Learning with SHAP and XGBoost for Lung Cancer Prediction Insights

Taufik Kurniawan — 2024-11-01

Lung cancer remains one of the leading causes of death worldwide, and early detection through accurate and reliable methods is essential to improve patient prognosis. This study proposes a lung cancer classification model that integrates XGBoost with SHapley Additive exPlanations (SHAP) and Random Over Sampling (ROS) techniques to address the data imbalance problem. Using hyperparameter optimization through Optuna, the resulting model demonstrated superior performance, with an average accuracy of 96.84%, precision of 99.23%, recall of 94.51%, F1-score of 96.74%, specificity of 99.17%, and AUC of 96.84% in a 10-fold cross-validation evaluation. SHAP analysis provided significant interpretability, identifying key features such as gender, smoking habits, and physical signs of yellow fingers as the factors that most influence the model's predictions. The results of this study indicate that the proposed model is not only accurate, but also interpretable, making a significant contribution to supporting better clinical decision making in lung cancer diagnosis.

Application of Machine Learning Algorithm for Osteoporosis Disease Prediction System

Rajendra Artanto Wiryawan Sujana — 2024-11-01

Osteoporosis is a condition characterized by decreased bone density, leading to fragile and easily fractured bones. This disease is a significant concern as it can cause disability, fractures, and death, particularly in the elderly population. Early detection of osteoporosis is crucial to prevent disease progression through timely interventions. This study aims to develop a machine learning-based prediction system capable of detecting osteoporosis using three different algorithms, Random Forest, Support Vector Machine (SVM), and Gradient Boosting. The study involves analyzing and comparing the performance of these algorithms based on evaluation metrics such as Accuracy, Precision, Recall, and F1 Score. The data used is processed in two formats, namely ordinal and one-hot encoding, to assess the impact of encoding techniques on model performance. The results show that the Gradient Boosting algorithm performs the best on both types of data, with the highest Accuracy of 91.07% on the one-hot encoded data. Meanwhile, SVM and Random Forest also demonstrate competitive performance but with slightly lower results. This study concludes that Gradient Boosting is the most effective algorithm for osteoporosis prediction in this research. These findings can serve as a foundation for further development in the early detection of osteoporosis and support more effective and efficient prevention and treatment efforts.

Implementation of Samba Server Using OpenVPN Based on Single Board Computer (SBC) for Private Cloud Storage

Dwi Bayu Putra Pamungkas — 2024-11-05

In the current digital era, people's need for data storage media that is practical and can be accessed at any time is increasing. This research aims to design and implement a practical cloud-based data storage system using a Raspberry Pi 4 Model B device using the Samba and OpenVPN applications. The system focuses on storing users' data (private cloud), which allows users to directly access files and data via a storage server. The method used in this research includes a literature review to support system development. Testing was carried out to evaluate the security of the system being built by comparing access to private cloud server services before and after using the OpenVPN application. Test results show that using the OpenVPN application increases the security of data exchange, with good encryption in communications between client and server. The resulting system runs according to the initial design and can function as a secure private cloud system. This research can contribute to the development of efficient and secure data storage solutions, as well as show the potential for using the Raspberry Pi as an energy and cost-saving personal cloud server device.

Implementation of IDS and IPS for Detecting and Preventing TCP Port Scanning and ICMP Flooding Attacks

Iqbal Maqdum Razzanda — 2024-11-05

The implementation of Intrusion Detection System (IDS) and Intrusion Prevention System (IPS) is a crucial step in maintaining network security. This research aims to test the effectiveness of IDS and IPS in detecting and preventing TCP port scanning attacks and ICMP flooding attacks and also providing real-time notifications using Telegram. The methodology used includes configuring a test environment that reflects real network scenarios, where various attacks are initiated to test the IDS and IPS responses. The experimental results show that IDS is able to detect suspicious activity with a high degree of accuracy, while IPS is effective in blocking identified attacks, thereby reducing potential damage to the system. Proper implementation of IDS and IPS can significantly improve network security by early detecting and preventing cyberattacks.

Application of MobileNetV2 and SVM Combination for Enhanced Accuracy in Pneumonia Classification

Eka Putra Agus Meindiawan — 2024-11-05

Pneumonia is a very common respiratory infection in low- and middle-income countries and is still a leading cause of death, especially among children under five years old. Modern technologies, such as machine learning, offer significant potential in improving the automatic detection of pneumonia through chest X-ray (CXR) image analysis. This study aims to develop a more accurate pneumonia diagnosis system by evaluating various feature extraction methods. CXR datasets of pneumonia patients were resized to 180x180 pixels and balanced using the SMOTE-Tomek technique. Three main approaches were investigated: direct classification using Support Vector Machine (SVM) on the SMOTE-Tomek balanced dataset, feature extraction using Sobel edge detection followed by SVM classification, and feature extraction using MobileNet-V2 followed by SVM classification. The results showed that Scheme 1 achieved 97% accuracy, Scheme 2 decreased to 95%, and Scheme 3 achieved the highest accuracy at 98%. The lower accuracy in Scheme 2 is due to the limitations of Sobel edge detection, which reduces the key features in the CXR image. On the other hand, the improvement in Scheme 3 is due to the effective feature extraction capability of MobileNet-V2. In conclusion, the choice of feature extraction method plays an important role in determining the accuracy of an automated diagnostic system. This study builds on existing research and is expected to make a significant contribution to the development of more accurate and efficient automated diagnostic systems, which can ultimately help reduce pneumonia-related mortality.

Donor Segmentation Analysis Using the RFM Model and K-Means Clustering to Optimize Fundraising Strategies

Rezki . — 2024-11-05

This study aims to segment donors using the Recency, Frequency, Monetary (RFM) model and the K-Means algorithm to optimize fundraising strategies. The RFM model is used to measure donor engagement through three dimensions: Recency (the last time a donation was made), Frequency (the frequency of donations), and Monetary (the amount of donations). By utilizing RFM scores, donors are then grouped using the K-means algorithm to generate more specific donor segments. This study was conducted using donation data from a non-profit organization, focusing on strategies to improve donor loyalty and donation frequency. The segmentation results identified several key segments, including Loyal Donors, New Donors, Potential Donors, and Low-Priority Donors. Each segment exhibits different donation behavior characteristics and requires a different strategic approach. The implementation of these segmentation results is expected to help the organization design more effective communication strategies and donation programs, as well as improve donor retention and lifetime value. Additionally, this study identifies the potential for enhancing the analytical model for broader applications in the future. This research contributes to non-profit organizations by offering a more efficient approach to managing donor relationships.

Chat GPT Impact Analysis on API Testing: A Controlled Experiment

Yehezkiel David Setiawan — 2024-11-05

This research examines the impact of ChatGPT as a learning aid for students in API testing. A controlled experiment compared two groups: one utilizing ChatGPT and the other relying on traditional documentation. The findings indicate that participants using ChatGPT scored significantly higher in both exam tests compared to the documentation group, despite taking longer to complete tasks. Statistical analysis using t-tests confirmed these differences as significant. Post-test surveys revealed an increase in participants confidence and effectiveness in understanding and using APIs after interacting with ChatGPT. However, potential downsides, such as over-reliance on ChatGPT and insufficient deep conceptual understanding, were also observed. The results suggest that while ChatGPT can greatly enhance the quality of learning and productivity in API-related tasks, users must balance AI assistance with independent problem-solving skills. This study underscores the potential of ChatGPT as a valuable educational tool, provided it is integrated thoughtfully into the learning process.

Designing an Chatbot with NLP Technology in a Website-Based New Student Admission Information System

Muhammad Fathan Fauzan — 2024-11-05

In the fast pace of digitalization, student admission information system websites face the challenge of providing responsive and quality services to applicants. One emerging solution is the use of chatbots, which enable automated interaction with customers. Technology continues to transform over time. At SMK Insan Teknologi (InTek), the service process is still manual, such as physical archives for student registration, incomplete information, and the absence of an official website. To improve administration and data access, a web-based information system is offered. While the Chatbot helps in interactive services and time efficiency to answer registrants' questions, NLP is used to make the conversation in the chat more natural and easy to understand by registrants. The results of testing the system show that the system functions properly in responding to messages sent through the chatbot on the website both from the message text according to the intent, as well as abstract text and not according to the pattern with an accuracy rate of 87,5%. It is hoped that this research can improve the quality of service and administrative efficiency at SMK Insan Teknologi and can be applied in other educational institutions.

Performance Comparison of Random Forest and Decision Tree Algorithms for Anomaly Detection in Networks

Rafiq Fajar Ramadhan — 2024-11-06

The increase in cyber attacks has made network security a very important focus in this digital era. This research compares the performance of two machine learning algorithms, that is Random Forest and Decision Tree for detecting anomalies in networks using the UNSW-NB15 datasets, which include various types of attacks such as DoS, Backdoor, Exploits and others which will be used to train and test both models. The data collection method, pre-processing, data splitting and modelling using SMOTE method to handle data imbalanced were applied in both algorithms and then evaluated using accuracy, precision, recall and f1-score metrics. From the study result, it can be conclude that the Decision Tree algorithm performs better in detecting anomalies in binary data with an accuracy of 99,71%. However, in multi-class data, Random Forest showed slightly better performance, though it required significantly more time for training and prediction. Despite the small difference in accuracy, Decision Tree demonstrated faster prediction times, making it more efficient for time-sensitive applications. This research concludes that while Random Forest provides higher accuracy for complex datasets, Decision Tree offers a more time-efficient solution with comparable accuracy.

Automatic Vegetable Watering System Using Fuzzy Logic with Integration of Soil Moisture, Rain Sensors, and RTC

Dallarizki Arginanta — 2024-11-12

Conventional vegetable watering often presents challenges, particularly in ensuring that plants receive adequate water without excessive manual intervention. This research proposes a solution in the form of an automatic watering system using fuzzy logic, which integrates soil moisture sensors, rain sensors, and an RTC (Real-Time Clock) for scheduling. The system is designed to replace manual watering methods with an automated process, thus improving the efficiency and effectiveness of vegetable cultivation. The developed device uses a soil moisture sensor to monitor soil conditions, a rain sensor to detect rainfall, and an RTC to determine the optimal watering times. The Arduino Uno acts as the main controller that activates the water pump via a relay driver based on data received from the sensors. Test results show that the system operates according to the established criteria, with a satisfactory accuracy level. The system successfully waters the plants at 07:00 WIB and 15:00 WIB, based on dry soil conditions and no rain. The trials showed that the device has an average soil moisture measurement error of 5%, and a time discrepancy of about 22 seconds on the RTC module. Each 1% increase in soil moisture requires approximately 1 second of watering duration. Watering times are adjusted to prevent the plants from drying out or dying, with a soil moisture threshold of below 40% set as the condition for requiring watering.

Prediction of Air Quality Index Using Ensemble Models

Theresia Herlina Rochadiani — 2024-11-12

The impact of air pollution on health is measured by the Air Quality Index (AQI). Accurate AQI prediction is essential for pollution reduction and public health recommendations. Traditional methods of monitoring air quality are inaccurate and time-consuming. This study uses IoT-based air quality data from Kampung Kalipaten, Tangerang to build an AQI prediction model with machine learning, specifically an ensemble model. Ensemble techniques such as bagging and boosting, which increase the reliability of predictions by reducing model bias and inconsistency, improve AQI prediction. Four ensemble models used in this study, they are Random Forest Regressor, Gradient Boosting Regressor, Adaboosting Regressor, and Bagging Regressor. As the evaluation, RMSE and R² metrics used. Random Forest Regressor perform the best with RMSE value of 0.6054 and R² value of 0.6271, although no significant differences of RMSE and R² value of the rest models.

Comparison of Hadoop Mapreduce and Apache Spark in Big Data Processing with Hgrid247-DE

Firmania Dwi Utami — 2024-11-12

In today’s rapidly evolving information technology landscape, managing and analyzing big data has become one of the most significant challenges. This paper explores the implementation of two major frameworks for big data processing: Hadoop MapReduce and Apache Spark. Both frameworks were tested in three scenarios sorting, summarizing, and grouping using HGrid247-DE as the primary tool for data processing. A diverse set of datasets sourced from Kaggle, ranging in size from 3 MB to 260 MB, was employed to evaluate the performance of each framework. The findings reveal that Apache Spark generally outperforms Hadoop MapReduce in terms of processing speed due to its in-memory data handling capabilities. However, Hadoop MapReduce proved to be more efficient in specific scenarios, particularly when dealing with smaller tasks or when memory resources are limited. This is largely because Apache Spark can experience overhead when initializing tasks for smaller jobs. Furthermore, Hadoop MapReduce's reliance on disk I/O makes it more suitable for tasks involving vast amounts of data that surpass available memory. In contrast, Spark excels in situations where quick iterative processing and real-time data analysis are essential. This study provides valuable insights into the strengths and limitations of each framework, offering guidance for practitioners and researchers when selecting the appropriate tool for specific big data processing requirements, particularly with respect to speed, memory usage, and task complexity.

Text Data Security Using LCG and CBC with Steganography Technique on Digital Image

Muhammad Wildan — 2024-11-12

This research proposes a text data security method using a combination of Linear Congruential Generator (LCG), Advanced Encryption Standard (AES) Cipher Block Chaining (CBC) mode, and Least Significant Bit (LSB) steganography technique on digital images. The message scrambling process using LCG produces ASCII characters as noise that is inserted in the original message. After that, the message is encrypted using AES-256 CBC to provide additional security. The encryption result is then hidden in the digital image through LSB steganography technique. Tests were conducted on images with JPEG and BMP formats to measure the visual quality after the data insertion process, as measured by PSNR (Peak Signal-to-Noise Ratio). The test results show a PSNR value of 56.60 dB for JPEG images and 70.84 dB for BMP images. In addition, the insertion process in JPEG images degrades the image quality, mainly due to lossy compression, compared to the lossless BMP format. This study concludes that the proposed combination of methods is effective in hiding messages in images, but is susceptible to compression on lossy formats such as JPEG. The use of lossless image formats such as BMP or PNG is recommended to maintain data integrity.

Analysis of Splicing Manipulation in Digital Images using Dyadic Wavelet Transform (DyWT) and Scale Invariant Feature Transform (SIFT) Methods

Zumratul Muhidin — 2024-11-12

In the digital age, image manipulation is common, often done before publication on social media. However, this can lead to negative impacts, including visual deception. This research aims to detect splicing type image manipulation using Dyadic Wavelet Transform (DyWT) and Scale Invariant Feature Transform (SIFT) methods. The process starts with image decomposition using DyWT to obtain LL sub-images, followed by local feature extraction using SIFT. An application built on desktop-based Matlab source was developed to detect splicing forgery in digital images. The test used 20 images, this image dataset was taken from canon 5d mark II camera and Vivo X80 mobile phone. Each 10 original images, and 10 edited images. These 10 original images are left as they are without making changes, editing or manipulation, while the other 10 images are changed, edited or manipulated using editing software, the results of this editing are uploaded to social media, such as Facebook and Instagram, which will later be used as datasets in testing. The results show that the splicing technique is detected accurately, and processing is faster on images with low pixel resolution. The DyWT and SIFT methods are effective in detecting post-processing attacks such as rotation and rescaling, although they have drawbacks. DyWT struggles in detecting subtle changes and noise, while SIFT is less effective on non-geometric manipulations. Overall, both methods face challenges in detecting complex manipulations and require significant computational resources, especially on high-resolution images.

A Comparison of Convolutional Neural Network (CNN) and Transfer Learning MobileNetV2 Performance on Spices Images Classification

Khoirizqi Velarati — 2024-11-12

This research was conducted to analyze the performance of the CNN algorithm without transfer learning in classifying spice images and compare it with the CNN algorithm using transfer learning on the MobileNetV2 architecture. This comparison aims to evaluate both methods' accuracy, efficiency, and overall performance and analyze the impact of transfer learning on classification results in the context of spices. The dataset consists of 1500 spice images divided into 10 classes, with each class of 150 images. In the first experiment, CNN without transfer learning resulted in 93% accuracy performance. For the second experiment using MobileNetV2, there was an increase in accuracy, reaching a value of 99% for all spice classes. The results of this study confirm that MobileNetV2 architecture significantly improves the accuracy and performance of spice classification compared to CNN without transfer learning, which can be recommended for spice image classification.

Implementation of the Naive Bayes Classifier Algorithm for Classifying Toddler Nutritional Status

Muhammad Insan Kamil — 2024-11-12

This research addresses the pressing issue of malnutrition among toddlers in Indonesia, aiming to classify their nutritional status using the Naive Bayes Classifier (NBC). The study utilizes a dataset comprising 958 records from Puskesmas Cilandak and categorizes nutritional status into six class labels: good nutrition, at risk of excess nutrition, excess nutrition, obesity, undernutrition, and severe malnutrition. The methodology includes data preprocessing techniques such as class weighting to tackle class imbalance and Principal Component Analysis (PCA) for effective feature extraction. The model's performance is evaluated using metrics such as accuracy, precision, recall, and F1 score, achieving an impressive accuracy of 85.76% when class weighting is applied, which significantly enhances the recall and F1 scores for minority classes. The findings highlight the critical importance of robust preprocessing and evaluation metrics in improving machine learning models for public health applications. Furthermore, they suggest that further exploration of alternative algorithms and dataset expansion could yield more comprehensive insights into the classification of toddler nutritional status.

Random Forest Algorithm for Toddler Nutritional Status Classification Website

Maylia Fatmawati — 2024-11-12

Accurate data processing is essential for classifying toddler nutritional status on a website platform. The Random Forest algorithm is particularly effective in this context due to its ability to manage large datasets and mitigate overfitting. This study leverages Flask as the web framework to ensure responsiveness and adaptability, optimizing the data processing experience for users. Using secondary data comprising 120,999 records, the research aims to answer: "What factors affect the accuracy of the Random Forest model in classifying toddler nutritional status?" Model evaluation yielded excellent performance metrics, with accuracy, precision, recall, and F1-score values of 99.91%, 100%, 100%, and 100%, respectively. These results highlight the informative attributes in the dataset, such as age, gender, and height, that enhance classification accuracy. The Flask-based website enables users, such as healthcare professionals and policymakers, to input essential data points and receive instant classification results, thereby supporting prompt and informed responses to nutritional health issues. This study confirms that the Random Forest algorithm, combined with an intuitive web interface, effectively classifies toddler nutritional status with high accuracy.

Utilization of ResNet Architecture and Transfer Learning Method in the Classification of Faces of Individuals with Down Syndrome

Made Doddy Adi Pranatha — 2024-11-13

Classifying the faces of individuals with Down Syndrome poses a significant challenge in image processing and genetic anomaly detection. This study leverages the ResNet34 architecture and transfer learning methods to improve classification accuracy for Down Syndrome facial recognition. Three experiments were conducted, varying the batch size, learning rate, and number of epochs. In the first experiment, the model achieved an accuracy of 82.83%, precision of 0.8362, recall of 0.8350, and an F1 score of 0.8348, showing promising performance but falling short of the target accuracy of 85%. The second experiment yielded the best results, with an accuracy of 87.88%, precision of 0.8956, recall of 0.8956, and an F1 score of 0.8956, indicating an optimal balance between correct predictions and errors. The third experiment resulted in the lowest accuracy, at 80.47%, with a precision of 0.8272, recall of 0.8249, and an F1 score of 0.8247, signifying a decline in performance compared to the other trials. Among the three experiments, the best configuration was achieved in the second trial, as the high recall value is crucial in medical contexts to ensure that as many individuals with Down Syndrome are correctly detected as possible, minimizing the risk of serious consequences due to false negatives.

Betta Fish Identification System Based On Convolutional Neural Network

Gilang Ardhi Saputra — 2024-11-13

This study developed an automated identification system based on the Convolutional Neural Network (CNN) to classify Betta Splendens, a fish species with high economic value in Indonesia. The system aims to improve accuracy and efficiency in the identification process. The research was divided into several experiments, where the data was split into 320 images for training, 80 for validation, and 100 for testing. We used two optimizers, Adam and RMSprop. The Adam optimizer experiments conducted two stages with learning rates of 0.0001 and 0.001, each with 100 and 200 epochs. The results showed that a lower learning rate (0.0001) with 200 epochs yielded the best test accuracy of 71%, while a learning rate of 0.001 caused accuracy to stagnate at 66%, indicating potential overfitting. The RMSprop optimizer with a learning rate of 0.00001 demonstrated good stability, though with slightly lower accuracy than Adam. This study highlights the importance of selecting the appropriate learning rate and number of epochs to achieve an optimal balance between training, validation, and testing accuracy, ensuring the model generalizes well to new data.

Comparison of EfficientNetB1 Model Effectiveness in Identifying Fish Diseases in South Asian Fish Diseases and Salmon Fish Diseases

Rahmanda Afridiansyah — 2024-11-13

The purpose of this study is to evaluate the effectiveness of the EfficientNetB1 model in identifying fish diseases across two distinct datasets: South Asian Fish Diseases and Salmon Fish Diseases. The South Asian Fish Diseases dataset includes seven categories: red bacterial disease, aeromoniasis, gill bacterial disease, fungal saprolegniasis, parasitic disease, and white tail viral disease. The Salmon dataset is divided into two parts: FreshFish and InfectedFish. Using the EfficientNetB1 algorithm, each dataset was separately trained and tested to predict species and disease. Results showed an accuracy of 98.14% for the South Asian Fish Diseases dataset and 99.18% for the Salmon Diseases dataset. These findings support the argument that the model possesses sufficient capability to detect diseases affecting various fish species. This suggests that the model could be a valuable tool in the aquaculture industry for disease management and detection strategies.

Real-time Detection Transformer (RT-DETR) of Ornamental Fish Diseases with YOLOv9 using CNN (Convolutional Neural Network) Algorithm

Dwi Nurul Huda — 2024-11-13

The lack of specialized tools to check the condition of ornamental fish has hindered effective management. This research proposes a novel software architecture that uses the YOLOv9 model combined with RT-DETR to enable accurate and timely identification of ornamental fish conditions including fish diseases, empowering farmers and hobbyists with a valuable resource. This integration is done using Soft Voting Ensemble Learning technique. To achieve this goal, an Android mobile application successfully classified healthy fish and accurately identified common diseases such as bacteria, fungal, parasitic, and whitetail. Based on the test results, the integration accuracy of the YOLOv9 and RT-DETR models produced a high result of 0.8947 while the stand-alone YOLOv9 showed 0.8889 and the stand-alone RT-DETR of 0.8904. Recommendations are given for the combination of YOLOv9 and RT-DETR in condition detection and diagnosis of ornamental fish diseases.

Comparison of Oversampling Techniques on Minority Data Using Imbalance Software Defect Prediction Dataset

Deni Hidayat — 2024-11-13

Software Defect Prediction Dataset as a component of the Software Defect Prediction model has a very vital role. However, NASA Software Defect Prediction has a problem with imbalance in minority data. This study compares the performance of oversampling techniques in overcoming this. A total of 90 oversampling techniques in the form of SMOTE and its variants were used. The results of this study indicate that there is no oversampling technique that is able to overcome this. The original dataset without oversampling shows good performance at the level of accuracy and f1-score but has low performance on auc-score and g-score. Several oversampling techniques show increased performance on auc-score and g-score, unfortunately at the same time showing a decrease in performance on accuracy and f1-score.

Evaluation of the Decision Tree Model for Air Condition Classification on the Global Air Pollution Dataset

Cindy Dinda Sabella — 2024-11-14

Air pollution is an urgent global environmental problem, with significant impacts on public health and ecosystem stability. This research aims to develop an air quality classification model using the Global Air Pollution dataset from Kaggle, which consists of 23,463 rows of data and 12 features, including important variables such as Air Quality Index (AQI), PM2.5, NO2, and O3. Decision Tree, Random Forest, and Support Vector Machine (SVM) algorithms are applied to perform classification, with a focus on hyperparameter tuning to increase model accuracy. The research results show that the Decision Tree provides the best results with an accuracy of 99.89% after tuning hyperparameters using the Grid Search method. The SVM model showed an improvement of 94.89% to 99.32%, while Random Forest recorded an accuracy of 96.87% with no significant improvement after tuning. Importance feature analysis identified PM2.5 and AQI as the dominant factors in influencing air quality, with PM2.5 having the highest importance value of 0.93. This research confirms that machine learning can be an effective tool for integrating and classifying air pollution. It is hoped that the integration of this model into a real-time air quality monitoring system can help make more responsive and precise decisions in dealing with air pollution problems.

Aspect-Based Sentiment Analysis for Enhanced Understanding of 'Kemenkeu' Tweets

Priska Trisna Sejati — 2024-11-14

The perceptions and expressions shared by the public on social media play a crucial role in shaping the reputation of government institutions, such as the Ministry of Finance MOF (Kemenkeu) in Indonesia which also has faced increased scrutiny, particularly on Twitter. This study analyzes public sentiment towards the Indonesian Ministry of Finance (MoF) through Aspect-Based Sentiment Analysis (ABSA) on Twitter data. Using a dataset of 10,099 tweets from January to July 2024, this study combines IndoBERT for sentiment classification and Latent Dirichlet Allocation (LDA) for topic modeling. Here, LDA was tested across four scenarios that considered various combinations of stopwords removal and stemming techniques, resulting in coherence scores of 0.314256, 0.369636, 0.350285, and 0.541752. The most optimal results were achieved in the scenario of stopwords removal without stemming (with 0.314256 coherence score). The main results show: 1) Identification of four main topics related to MoF: Economy, Budget, Employees, and Tax; 2) The dominance of negative sentiment (6,837 tweets) compared to positive sentiment (198 tweets) across all topics; 3) The effectiveness of IndoBERT in handling the complexity of the Indonesian language, especially in interpreting context and language nuances; 4) The importance of proper preprocessing, with a scenario of removing stopwords without stemming resulting in the most relevant topics. This study provides valuable insights for MoF to understand public perception and identify areas that require special attention in public communication and policy.

Optimization of Tourism Destination Recommendations in Batang Regency Using Content-Based Filtering

Ilmira Yulfihani — 2024-11-15

In an era where tourism plays a pivotal role in economic development, the need for effective navigation through diverse attractions has never been more critical. This research presents a cutting-edge tourism recommendation system tailored for Batang Regency, leveraging Content-Based Filtering (CBF) to deliver personalized suggestions that enhance the tourist experience. By categorizing tourist attractions into Culinary, Culture, Accommodation, Nature, and Leisure, and employing the Haversine formula for precise geographical calculations, our system prioritizes recommendations based on user preferences and proximity. Recommendation testing yielded an impressive average F1 Score of 0.965, underscoring the system's accuracy and relevance, particularly in straightforward user scenarios. However, the research also identifies challenges in more complex cases, suggesting the need for future enhancements through hybrid models and the integration of user feedback. This innovative approach not only streamlines the decision-making process for tourists but also aims to boost local tourism, making it an invaluable tool for both visitors and the Batang Regency community. Join us in exploring how technology can transform the way we experience travel, ensuring that every journey is tailored to individual desires and needs.

Comparison of Clustering Algorithms: Fuzzy C-Means, K-Means, and DBSCAN for House Classification Based on Specifications and Price

Dhendy Mardiansyah Putra — 2024-11-15

This study aims to compare the performance of three clustering algorithms, namely Fuzzy C-Means, K-Means, and DBSCAN, in grouping houses based on their specifications and prices. The data used includes features such as price, building area, land area, number of bedrooms, number of bathrooms, and availability of garages. The performance of these algorithms was evaluated using Silhouette Score and Davies-Bouldin Score to determine the quality of cluster separation. The results indicate that K-Means achieved the best performance with the highest Silhouette Score of 0.7702 for two clusters, followed by Fuzzy C-Means, which excelled in handling overlapping clusters. DBSCAN, while effective in detecting outliers, showed suboptimal performance for this housing dataset. These findings suggest that K-Means is the most suitable clustering method for housing data, while Fuzzy C-Means and DBSCAN can serve as alternatives depending on the data characteristics. This research is expected to assist in making the house searching and classification process more efficient and provide additional insights for developers in shaping housing market strategies.

Sentiment Analysis of Online Vehicle Tax Renewal Application Users Using Support Vector Machine Algorithm

Muhamad Ilham Fauzy — 2024-11-15

This study examines user sentiment towards online vehicle tax renewal applications by utilizing the Support Vector Machine (SVM) algorithm. The data was collected from user reviews on the Google Play Store for three major applications: New Sakpole, Sapawarga, and Timsalut. The reviews were preprocessed through steps including normalization, case folding, tokenization, and stopword removal. The SVM algorithm was then applied to classify the reviews into positive or negative sentiments. A comparative analysis was performed with K-Nearest Neighbors (KNN) and Naïve Bayes, with SVM demonstrating the best performance, achieving an accuracy of 76.5%. In addition to accuracy, metrics such as precision, recall, and F1-score were also evaluated to provide a more comprehensive assessment of the models. The results indicate that while these applications help facilitate vehicle tax payments, there remains significant user dissatisfaction, particularly related to technical issues and usability concerns. This study offers valuable insights for application developers, highlighting areas for improvement in functionality and user experience to better meet public expectations.

Comparison of Naïve Bayes Classifier and Decision Tree Algorithms for Sentiment Analysis on the House of Representatives' Right of Inquiry on Twitter

Putri Wahyuni — 2024-11-20

This research analyzes public sentiment towards the topic of the House of Representatives' Right of Inquiry on Twitter using Naïve Bayes Classifier and Decision Tree algorithms. The goal is to compare the effectiveness of the two algorithms in political sentiment analysis. . The research methodology includes data collection from Twitter, data pre-processing, sentiment classification, and result analysis. Sentiment analysis reveals the dominance of positive sentiment related to the DPR's Right of Inquiry. However, this study has limitations in terms of dataset size and depth of text-based sentiment analysis. This research contributes to a better understanding of public sentiment towards political issues in Indonesia and highlights the importance of proper algorithm selection in social media sentiment analysis. Development suggestions include exploration of deep learning techniques, integration of multimodal analysis, data balancing (oversampling or undersampling) and improvement of pre-processing so that the model is better able to capture negative contexts. The results of the study showed excellent performance of both Naive Bayes Classifier and Decision Tree algorithms with accuracy above 95%. Decision Tree excels with an accuracy of 99%, while Naïve Bayes Classifier performs better with an accuracy of 96%. The results with the Confusion Matrix test are precision 0.98, recall 1.00, and F1-Score 0.99.

Sentiment Analysis on Tabungan Perumahan Rakyat (TAPERA) Program by using Support Vector Machine (SVM)

Rizki Agam Syahputra — 2024-11-20

This study aims to analyze public sentiment towards the Housing Savings Program (TAPERA) using the Support Vector Machine (SVM) algorithm. The dataset comprises 16,061 reviews about TAPERA which was gathered from web scrapping and YouTube API. The sentiment analysis results indicate that 99.8% of the reviews are negative, while only 0.2% are positive. The SVM model applied in this study achieved a very high accuracy rate of 99.81%. This indicates that the model is highly effective in classifying sentiments, particularly in identifying negative sentiments. The resulting confusion matrix shows the model's excellent performance in detecting negative sentiments, with no False Positives (FP) and a very high number of True Negatives (TN). However, the model exhibits weaknesses in detecting positive sentiments, as indicated by the presence of several False Negatives (FN) and the absence of True Positives (TP). The findings of this study suggest that the public generally holds a very negative view of the TAPERA program. This insight is crucial for program administrators to consider as they evaluate and improve the program based on negative feedback received from the public. Overall, this research provides important insights into public perceptions of TAPERA and underscores the need for better modeling for more representative sentiment analysis. These findings can serve as a basis for policymakers in designing more effective communication strategies and program improvements to increase public acceptance of TAPERA.

Sentiment Analysis of Indonesian Responses to the Conflict in Palestine Using KNN and SVM Methods

Rizky Fauzi — 2024-11-20

The prolonged conflict between Palestine and Israel has attracted worldwide attention, including Indonesia, which has a history of strong support for the Palestinian cause. This study aims to analyze the sentiment of Indonesian people towards the Palestinian-Israeli conflict using the K-Nearest Neighbors (KNN) and Support Vector Machine (SVM) methods. The subject of this research is user data X (Twitter) which contains opinions about the conflict. After preprocessing, weighting, and labeling, 2960 tweets were collected and classified into three sentiment categories: positive, negative, and neutral. The KNN+SVM method is applied to classify the sentiment of the processed tweet data. The results showed that of the 2960 data analyzed, 33.8% were labeled positive, 38.9% were labeled negative, and 27.4% were labeled neutral with 82% accuracy, 83% precision, 82% recall, and 82% F1-Score. These results show that the majority of Indonesians tend to be negative in expressing their views on the Palestinian-Israeli conflict. This analysis provides greater insight into sentiment patterns in Indonesian responses to sensitive issues, and contributes to the study of public opinion and social dynamics on social media.

Detecting Fake Reviews Using BERT and Sublinear_TF Methods on Hotel Reviews in the Lombok Tourism Area

Zulpan Hadi — 2024-11-20

The number of visitors to Lombok, one of the famous tourist destinations in Indonesia, increased from 400,595 in 2020 to 1,376,295 in 2022. Although the government supports the hotel industry, fake reviews are a significant problem that can damage hotel reputations and mislead tourists. This study uses BERT and Sublinear_TF feature extraction techniques to analyze fake reviews from three main areas: Gili Trawangan, Senggigi, and Kuta. BERT detects fake reviews by understanding the context of words, while Sublinear_TF emphasizes more informative words by reducing the weight of irrelevant common words. The results showed that the more extensive and diverse dataset from Gili Trawangan had the best classification results. The combination of BERT and Random Forest achieved the highest accuracy of 0.84. Overall, BERT excels in Gili Trawangan with an accuracy of 0.79 for SVM and 0.84 for Random Forest. In contrast, smaller and more homogeneous datasets such as Senggigi and Kuta have lower accuracy. In addition, Sublinear_TF performed well on Gili Trawangan with an accuracy of 0.82 using SVM and 0.83 using Random Forest; however, its performance declined in Senggigi and Kuta. BERT and Sublinear_TF techniques are more effective on large and diverse datasets such as Gili Trawangan. Sublinear_TF is better for varied data but less effective on more homogeneous datasets, while BERT with Random Forest showed the highest accuracy due to its ability to capture broader language context. This suggests that the size and variety of the dataset highly influence the success of fake review classification techniques.

Comparative Study: Flower Classification using Deep Learning, SMOTE and Fine-Tuning

Vincentius Praskatama — 2024-11-20

Deep learning is a technology that can be used to classify flowers. In this research, flower type classification using the CNN method with several existing CNN architectures will be discussed. The data consists of 4317 images in .jpg format, covering 5 classes that is sunflower, dandelion, daisy, tulip and rose. The distribution of data for each class is daisy with 764 pictures, dandelion with 1052 pictures, rose with 784 pictures, sunflower with 733 pictures, and tulip with 984 pictures. With total dataset of 4317 pictures is further split to training data with ratio of 60%, validation with ratio of 10%, and testing with ratio of 30% to process with the CNN method and CNN framework. Due to the imbalance data distribution, the SMOTE method is applied to balancing number of samples in each class. This research compares CNN architectures, including CNN, GoogleNet, DenseNet, and MobileNet, where each transfer learning model undergoes fine-tuning to improve performance. At the classification stage, performance will be measured based on model testing accuracy. The accuracy obtained using CNN is 74.61%, using GoogleNet is 87.45%, DenseNet is 93.92%, and MobileNet is 88.34%.

ROVIGA: Model-Driven Soil Moisture Sensor for Internet-Connected Plant Pot

Iman Setiawan — 2024-11-20

The soil moisture sensor provides numerical measurements to detect changes in soil moisture using an analog voltage output. This research aims to develop a capacitive sensor based on a statistical model to detect soil moisture for plant watering, leveraging the Internet of Things (IoT). The analysis was conducted using polynomial and linear regression models. The modeling process was based on primary gravimetric test results from dried soil. The best model coefficients, selected based on the highest adjusted R-squared value, were used for sensor recalibration. A watering system was then developed using an Arduino and a model-driven capacitive soil moisture sensor integrated into an internet-connected smart plant pot, enabling remote control via a mobile phone. The research findings indicate that the 8th-order polynomial model, with the highest adjusted R-squared value of 0.9583, is the most accurate. The smart watering system using the model-driven capacitive sensor achieved soil moisture prediction outcomes ranging from 0.08 to 1.01 for 150 to 418 sensor data points. The internet-connected smart plant pot allows precise and real-time control, delivering notifications and enabling actions when plants require watering.

Implementation of AlexNet and Xception Architectures for Disease Detection in Orange Plants

Venus Al Fatah — 2024-11-20

Oranges are one of Indonesia's primary horticultural commodities, with production increasing each year. However, pest and disease infestations often go undetected, leading to significant reductions in crop yields. This study implements Convolutional Neural Network (CNN) technology to identify diseases in orange plants using two architectures: AlexNet and Xception. The implementation results show that the Xception architecture achieved a high accuracy of 96% after 100 training epochs, indicating its effectiveness in disease detection tasks. This research highlights the potential of integrating CNN technology, particularly the Xception model, into web-based systems for disease detection in orange plants. Such systems can assist farmers in maintaining crop health, improving productivity, and ensuring harvest quality.