Journal of Applied Informatics and Computing

Crack Detection in Building Through Deep Learning Feature Extraction and Machine Learning Approch

2024-07-02T09:48:51+00:00

Buildings with cracks are extremely hazardous because they have the potential to cause destruction. Numerous occupants of structures such as houses and buildings are at risk when cracks appear. There are numerous techniques for identifying fractures in structures, including visual inspection, tool use, and expert inspection. The present study employed computer vision, a form of artificial intelligence, to detect cracks in buildings. The main objective of this research is to construct a prototype capable of real-time monitoring of cracks in building walls. This research makes use of a methodology that combines machine learning and deep learning. Machine learning is employed in the classification process, whereas deep learning is utilized for the extraction of features. This research employs MobileNetV2 as its deep learning architecture and K-NN, Naive Bayes, SVM, XGBoost, and Random Forest as its machine learning classifiers. Test results show that when dividing the 80:20 dataset, XGBoost algorithms can produce the highest accuracy, sensitivity, and specificity values of 99%. Tests in the real environment are performed by deploying Raspberry Pi. Test results show that the prototype can detect cracks inthe structure surfaceat a distance of 10 meters in a bright environment. The crack detection process is carried out in real time at an average speed of 42fps.

IndoBERT Model Analysis: Twitter Sentiments on Indonesia's 2024 Presidential Election

2024-07-02T09:48:52+00:00

Elections are one of the key moments in a country's democracy. Indonesian elections have a significant impact on regional and global politics. Twitter being one of the popular social media platforms becomes a powerful tool for political campaigns. This makes it an ideal source to analyze public opinion during the 2024 general election, particularly the upcoming Presidential Election (Pilpres). IndoBERT is the model chosen to analyse the sentiment from the dataset in this study using a zero-shot learning approach. Based on the evaluation results, the accuracy value of the 2024 presidential election classification is 0.60 (60%), tends to predict with a good value in the positive label of 0.74 (74%) for F1-Score. This model is considered quite good at predicting negative labels but the results are not too optimal with a value of 0.49 (49%). Confusion Matrix in this IndoBERT model is more likely to label tweets with positive things, by detecting negative labels quite well.

Optimization Chatbot Services Based on DNN-Bert for Mental Health of University Students

2024-07-02T09:48:53+00:00

Attention to mental health is increasing in Indonesia, especially with the recent increase in the number of cases of stress and suicide among students. Therefore, this research aims to provide a solution to overcome mental health problems by introducing a chatbot system based on Deep Neural Networks (DNN) and BiDirectional Encoder Representation Transformers (BERT). The primary objective is to enhance accessibility and offer a more effective solution concerning the mental health of students. This chatbot utilizes Natural Language Processing (NLP) and Deep Learning to provide appropriate responses to mild mental health issues. The dataset, comprising objectives, tags, patterns, and responses, underwent processing using Indonesian language rules within NLP. Subsequently, the system was trained and tested using the DNN model for classification, integrated with the TokenSimilarity model to identify word similarities. Experimental results indicate that the DNN model yielded the best outcomes, with a training accuracy of 98.97%, validation accuracy of 71.74%, and testing accuracy of 71.73%. Integration with the TokenSimilarity model enhanced the responses provided by the chatbot. TokenSimilarity searches for input similarities from users within the knowledge generated from the training data. If the similarity is high, the input is then processed by the DNN model to provide the chatbot response. This integration of both models has proven to enhance the responsiveness of the chatbot in providing various responses even when the user inputs remain the same. The chatbot also demonstrates the capability to recognize various inputs more effectively with similar intentions or purposes. Additionally, the chatbot exhibits the ability to comprehend user inputs although there are many writing errors.

Forecasting Air Quality Indeks Using Long Short Term Memory

2024-07-02T09:48:54+00:00

Exercise offers significant physical and mental health benefits. However, undetected air pollution can have a negative impact on individual health, especially lung health when doing physical activity in crowded sports venues. This study addresses the need for accurate air quality predictions in such environments. Using the Long Short-Term Memory (LSTM) method or what is known as high performance time series prediction, this research focuses on forecasting the Air Quality Index (AQI) around crowded sports venues and its supporting parameters such as ozone gas, carbon dioxide, etc. -others as internal factors, without involving external factors causing the increase in AQI. Preprocessing of the data involves removing zero values and calculating correlations with AQI and the final step performs calculations with the LSTM model. The LSTM model which adds tuning parameters, namely with epoch 100, learning rate with a value of 0.001, and batch size with a value of 64, consistently shows a reduction in losses. The best results from the AQI, PM2.5, and PM10 features based on performance are MSE with the smallest value of 6.045, RMSE with the smallest value of 4.283, and MAE with a value of 2.757.

Enterprise Architecture Model of the New Student Admission System at Stella Maris University Sumba

2024-07-02T09:48:55+00:00

This research aims to design an Enterprise Architecture (EA) model for the new student admission system at Stella Maris Sumba University. The background of this research is the need to improve efficiency, transparency, and integration in the new student admission process, which currently still faces various administrative and technical challenges. The research method used is qualitative which includes literature studies and in-depth interviews with relevant parties. The data obtained was analyzed to identify needs and design the right EA model. The purpose of this research is to create a system capable of automating the admission workflow, ensuring data security, and providing real-time access for application status tracking. The results showed that the proposed Enterprise Architecture model can improve operational efficiency, user satisfaction, and support the strategic decisions of Stella Maris Sumba University based on accurate and integrated data.

Comparison of Deep Learning Architectures in Identifying Types of Medicinal Plant Leaf Images

2024-07-02T09:48:56+00:00

This study focuses on the identification of 3500 images of medicinal plant leaves using Deep Learning CNN Transfer Learning models such as MobileNet, VGG16, DenseNet121, ResNet50V2, and NASNetMobile. The dataset used is the "Indonesian Herb Leaf Dataset 3500," consisting of 10 classes of medicinal plants. This research has the potential to efficiently and accurately recognize medicinal plants using machine learning workflow methods. The objective of this study is to compare the performance of these five methods in conducting plant identification. The testing phase involves various data handling schemes, dividing the data into two scenarios: 80:10:10 and 70:20:10. Performance comparison is also done between augmented and non-augmented data. The research findings indicate that MobileNet exhibits the best performance with an accuracy, precision, recall, and f1-Score of 98.86%. Accurate leaf identification supports further research on the properties and benefits of medicinal plants and can be applied in the development of decision support systems for plant recognition.

Taxpayer Awareness Classification Using Decision Tree and Naïve Bayes Methods

2024-07-02T09:48:57+00:00

Land and Building Tax (PBB) has a big influence on a region's PAD. Therefore, regions always strive to increase PBB income as much as possible. Many factors influence the increase in PBB, one of which is public awareness of taxes. Lack of public awareness of taxes causes PBB income to also decrease, and has implications for regional PAD. And conversely, if public awareness of taxes is high, PBB and PAD revenues will also increase. Therefore, a system is needed to measure public awareness of taxes in the region. If public awareness of taxes can be measured, then the relevant agencies can evaluate and map taxpayers in which sub-districts have high or below average awareness. There are several factors that influence taxpayer awareness, including ownership status, tax sector, assessment category, and the number of receivable payments over the past 5 years. By knowing the awareness of taxpayers, the relevant agencies can review the targets for achieving PBB revenue and issue warning letters to taxpayers whose awareness of PBB is lacking. This research uses decision tree and naïve Bayes methods to classify 666,580 datasets obtained from the Cianjur Regency Regional Revenue Management Agency. The stages are carried out by data collection, data preprocessing, training data labeling, classification process, and evaluation. The result of this research is a system that can predict whether taxpayers are aware or not in a sub-district and sub-district or rural area using decision trees and naïve Bayes. The accuracy obtained from the decision tree method was 93.73%, while the accuracy obtained from the naïve Bayes method was 85.61%.

Stock Market Index Prediction using Bi-directional Long Short-Term Memory

2024-07-02T09:48:59+00:00

The IHSG (Indonesia Stock Exchange Composite Index) is a stock price index in the Indonesia Stock Exchange (BEI) that serves as an indicator reflecting the performance of company stocks through stock price movements. Therefore, IHSG becomes a reference for investors in making investment decisions. Advanced stock exchanges generally have a strong influence on other stock exchanges. Several studies have proven the influence of one global index on another. Global index is a term that refers to each country's index to represent the movement of its country's stock performance. Forecasting IHSG can be one of the analyses that help investors make wise decisions when investing. To obtain an IHSG forecasting model, an appropriate and suitable method is needed, especially for data that has a large amount. LSTM is a development of Recurrent Neural Network (RNN) which has the ability to remember information in a longer period of time, while Bi-LSTM is a development of LSTM which has the ability to remember information longer and can understand more complex patterns than LSTM. This research provide the IHSG forecasting based on global index factors. The results showed that the best Bi-LSTM model (6-9-1) had a better performance in predicting and forecasting JCI movements with a MAPE value of 0.572314% better than the best LSTM model (4-10-1) which had a MAPE of 0.74326%. With forecasting based on the Bi-LSTM model, it is expected to help investors in making decisions on the Indonesia Stock Exchange (IDX).

Implementation of K-Means, Hierarchical, and BIRCH Clustering Algorithms to Determine Marketing Targets for Vape Sales in Indonesia

2024-07-02T09:49:01+00:00

In today's era, smoking is a common thing in everyday life. Along with the development of the times, an innovation emerged, namely the electric cigarette or vape. Electric cigarettes or vapes use electricity to produce vapor. The e-cigarette business is very promising in today's business world due to the consistent increase in market demand. However, determining the target buyer is one of the things that is quite important in determining the success of a business. In this analysis, the background of each region in Indonesia has different diversity; therefore, observation of data is needed to find out which regions in Indonesia have the potential to increase marketing based on profits (margins) to support the target market analysis process so that companies do not suffer losses and increase business success. In this study, the analysis will be carried out using vape quantity, margin, and purchasing power data in each region, which is processed using 3 algorithms: K-Means, Hierarchical, and BIRCH. The results of the clustering of the three algorithms produce two clusters. The K-means, Hierarchical, and BIRCH algorithms produce the same clusters: a potential cluster consisting of 18 cities and a non-potential cluster consisting of 45 cities. To see the performance of the model results, an evaluation was carried out using the Silhouette score, Davies Bouldin, Calinski Harabasz, and Dunn index, which obtained results of 0.765201, 0.376322, 315.949434, and 0.013554. From these results, it can be concluded that the clustering results are not too good and not too bad because the greater the Silhouette Score, Calinski Harabasz, and Dunn Index value, the better the clustering results while for Davies Bouldin the smaller the value means the better the clustering results.

Sentiment Analysis of Sirekap Application Users Using the Support Vector Machine Algorithm

2024-07-02T09:49:03+00:00

In the current era of digitalization, various activities are conducted using technology to aid their execution, including the democratic process scheduled for February 2024. The Komisi Pemilihan Umum (KPU) is utilizing a mobile-based application called Sirekap. During the previous presidential and vice-presidential elections, there were many pros and cons regarding the Sirekap application. A significant number of negative reviews were expressed by the public towards this application. This study employs the SVM algorithm to perform sentiment analysis of Sirekap application users. Before building the model, several steps were undertaken, including data labeling, data preprocessing, splitting the dataset into training and testing data, and performing transformations using Count Vectorizer. Evaluation of the SVM model results shows quite good performance with an accuracy of 81%. For the negative class, the precision and recall values are 87% and 85%, respectively, while for the positive class, the precision and recall values only reach 66% and 70%, indicating a need for improvement in the model's identification of the positive class. Five-fold cross-validation was performed with an average cross-validation score of 79.6% and a standard deviation of 2.14%, indicating the model's consistency across various training data subsets. These findings suggest that the SVM model can effectively perform text classification tasks. Based on the negative word cloud, it can be concluded that the Sirekap application still has many shortcomings affecting the democratic process in February 2024.

Analysis of pH and Turbidity Sensor Outputs in Shrimp Ponds for Vannamei Shrimp Commodities

2024-07-02T09:49:04+00:00

Vannamei shrimp is a high-value, economically important brackish water aquaculture commodity that is easy to cultivate. Optimal growth of Vannamei shrimp can be achieved by monitoring water quality parameters such as pH and turbidity. The pH levels can be measured using a pH sensor, with a pH meter as a reference. Turbidity levels are measured with a turbidity sensor in NTU units, with a turbidity stick serving as a reference. Testing of these sensors was conducted from morning to noon over three days in the brackish water ponds of IBAP Banjar Kemuning, Sidoarjo, recording 100 data samples. The performance of both sensors fluctuated due to disturbances around the pond, prompting the use of the Moving Average (MA) filter method to improve accuracy. Applying MA with varying window sizes (wz) resulted in a performance increase of 0.24% in the morning and 0.1% at noon. Additionally, turbidity sensor testing indicated that the pond conditions were consistent with the turbidity measurements obtained using the turbidity stick.

Text Insertion and Encryption Using The Bit-Swapping Method in Digital Images

2024-07-02T09:49:06+00:00

Communication is an essential aspect of everyday life, involving the transmission of information through various media. Technological advances have made communication easier but have also increased privacy and data security risks. Several efforts are made to maintain the security of digital information, including coding information (cryptography) and hiding information (steganography). In this article, the author secures information through a combination of cryptography and steganography. To secure text data, we encrypt by exchanging bits between adjacent characters. Subsequently, the encrypted text is hidden within an image. The security analysis results show the successful reconstruction of the message from the stego image and the successful restoration of the message to its original form. The use of the bit swapping method in the text message encryption process has been proven to enhance the security level of the ciphertext, as indicated by the lower TPK calculation value of 0.33 compared to the TPK value in previous studies. Additionally, embedding the ciphertext into digital images has been demonstrated to further increase the security level of the message, evidenced by the NPCR calculation value of 0.0000109% and the UACI calculation value of 0.000000555%. These very small values indicate no significant changes.

Implementation of Identity Loss Function on Face Recognition of Low-Resolution Faces With Light CNN Architecture

2024-07-02T09:49:07+00:00

Face recognition in low-resolution images has seen significant advancements over the past few decades. Although extensive research has been conducted to improve accuracy in these conditions, one of the main challenges remains the difficulty in identifying unique facial features in low-resolution images, leading to high error rates in identification. The use of Deep Convolutional Neural Networks (DCNN) for low-resolution face recognition is still limited. However, employing super-resolution models like REAL-ESRGAN can enhance recognition accuracy in low-resolution images. This study utilizes the Light CNN architecture and applies the margin-based identity loss function AdaFace on low-resolution datasets. The model is trained using the Casia-WebFace dataset and evaluated using the LFW and TinyFace test datasets. Based on the evaluation results on the LFW test data, the best model is Light CNN9-AdaFace, achieving the highest accuracy of 97.78% at 128x128 resolution. For images with the lowest resolution of 16x16, an accuracy of 83.37% was achieved using super-resolution techniques. On the TinyFace test data, the use of super-resolution resulted in performance metrics with a Rank-1 accuracy of 47.26%, Rank-5 accuracy of 55.25%, Rank-10 accuracy of 58.61%, and Rank-20 accuracy of 61.90% using the Light CNN9-AdaFace architecture.

Deployment of Kidney Tumor Disease Object Detection Using CT-Scan with YOLOv5

2024-07-02T09:49:09+00:00

Image processing plays a crucial role in identifying kidney tumors through CT-Scan images. Object detection technology, particularly YOLO, stands out for its speed and accuracy in facilitating more detailed analysis. Using Flask as a web framework offers optimal responsiveness, providing adaptive ease of use, especially in medical image processing. Evaluation of the model shows impressive results, with a mean Average Precision (mAP) of 0.987 for the 'kidney tumor' label. Detection on public data demonstrated high performance with accuracy, precision, recall, and F1-Score of 98.56%, 98.66%, 99.66%, and 99.16%, respectively. This study also utilized clinical data comprising 62 CT-Scan images. Evaluation of the clinical data revealed that YOLOv5 produced an accurate detection model with accuracy, precision, recall, and F1-Score of 95.16%, 96.72%, 98.33%, and 97.52%, respectively. The research shows that both public and clinical data models can accurately detect kidney tumors based on CT-Scan images. The deployment process using the Flask web-based platform allows direct interaction with users through an intuitive interface, enabling users to upload their CT-Scan images and quickly obtain detection results. These test results provide evidence that object detection using YOLOv5 achieves high accuracy in detecting both public and clinical datasets.

Web-Based Mapping of Crime-Prone Areas in Samarinda Seberang and Loa Janan Ilir Districts, Samarinda Citys

2024-07-02T09:49:10+00:00

The development of Geographic Information System (GIS) technology has provided significant benefits in various fields, including the monitoring of crime-prone areas. GIS is used to minimize the traces of these crimes. This study aims to map crime-prone areas in the Samarinda Seberang and Loa Janan Ilir Districts to identify which areas are potentially vulnerable, allowing for analysis for prevention and handling. The data used were collected from theft cases that occurred in these districts in 2019 and 2020. The research employs a scoring technique where each parameter is rated according to its classification. The results of the scoring process are then analyzed to determine the level of crime-prone areas, categorizing them as very vulnerable, vulnerable, or not vulnerable. Based on respondents' feedback, the application facilitates users in locating crime-prone areas, with 94.34% of responses indicating agreement or strong agreement. These results suggest that the application is feasible for implementation.

Predictive Analytics for IMDb Top TV Ratings: A Linear Regression Approach to the Data of Top 250 IMDb TV Shows

2024-07-02T09:49:12+00:00

In the era of a growing entertainment industry, understanding audience preferences and predicting the financial performance of entertainment products such as films and television shows has become increasingly important. Previous research has demonstrated various approaches in understanding the factors that influence the financial performance of entertainment products. However, there is still a need for research to investigate other aspects of film and television show evaluation. This study aims to explore the contribution of linear regression in analysing the ratings and financial performance of IMDb's top TV shows. Through the incorporation of various data-informed and interpretative approaches, it is expected to gain a deeper understanding of the factors that influence the success of a television show. Using data from the Top 250 IMDb TV Shows, a predictive analysis was conducted to understand the relationship between the number of episodes and IMDb ratings. The results of the information showed a negative relationship between the number of episodes and IMDb rating, with the linear regression model predicting a decrease in IMDb rating as the number of episodes increases. Implications of this research include recommendations for content creators to consider both quality and quantity of content in the development of TV shows.

Tourist Perceptions Through Sentiment Analysis to Support Tourism Development in Maluku Province

2024-07-02T09:49:14+00:00

Tourist perceptions obtained by sentiment analysis can provide an overview of tourism development in Maluku Province. This study aims to determine the perception of tourists towards destinations in Maluku based on the results of sentiment analysis. This research uses a quantitative approach by analyzing scrapping and snipping data from Facebook, Instagram, TikTok, Google Maps Review, and Trip Advisor. Sentiment analysis is done by comparing the accuracy level of the Random Forest, Naïve Bayes, and Support Vector Machine classification models. The results of the comparison of the three methods show that Random Forest has the best accuracy rate, which is 85%. The results of sentiment analysis both on the entire dataset and the results of analysis per district/city show that tourists' perceptions of tourist destinations in Maluku can be said to be good because they are dominated by negative sentiments. The existence of negative and neutral sentiments indicates that there is a need for improvement and improvement in the quality of tourist services in terms of human resources, transportation, accommodation, and infrastructure facilities.

Analysis of User Experience in the Design of the AMGM Lab Mobile Application Using the User Experience Questionnaire (UEQ) for Enhanced Efficiency

2024-07-02T09:49:15+00:00

AMGM mobile Lab application is a design application supporting documentation of water sample management data. There is a management support system in the form of an intranet. Nevertheless, access to the system is restricted to the office. Naturally, this is less effective since officers must first return to the office in order to enter sample data, and this cannot be done in real-time while they are out in the field. Thus, in order to facilitate these operations, mobile application development is required. User experience analysis is necessary to provide an inventive UX design, which is then required to satisfy laboratory requirements and the company's expectations for the application design. The objective of this study is to use user experience research as a foundation for future application development. In order to do user experience analysis, the User Experience Questionnaire (UEQ) approach is used. The test results reveal that the assessment falls into the excellent category for attractiveness (1.86) and dependability (1.82). The efficiency (1.85) and novelty (1.24) scales are categorized as good. The perspicuity (1.71) and stimulation (1.21) measures are categorized as above average. The mean of the entire scale exceeds 0.8, indicating that people assess all features positively.

Classification of COVID-19 Aid Recipients in Kasomalang District Using the K-Nearest Neighbor Method

2024-07-02T09:49:17+00:00

The impact of the Coronavirus, also known as COVID-19, which emerged in 2019, has not only threatened public health but also affected the global economy, including Indonesia. The government has initiated various aid programs to assist the community during the COVID-19 pandemic. These aids are expected to alleviate the economic burden on the affected population. One such aid program is the Direct Cash Assistance (Bantuan Langsung Tunai/BLT) from the Village Fund, which has been distributed since the onset of COVID-19 in Indonesia. However, the distribution of BLT has encountered several issues, including misidentification of recipients and double or excessive distribution beyond the established criteria. To address these issues, data mining for the classification of aid recipients can be employed. This study uses the K-Nearest Neighbor (KNN) method for data mining classification to classify residents' data with new patterns, ensuring aid distribution aligns with the criteria and eliminating double recipients. The application of K-Nearest Neighbor to the population data in Kasomalang District yields optimal performance, with evaluation results showing an accuracy of 96%, precision of 0.98, recall of 0.96, and F1 score of 0.97 using the confusion matrix method.

Applying the Multi-Attribute Utility Theory (MAUT) to Accurately Determine Stunting Susceptibility Levels in Toddlers

2024-07-02T09:49:18+00:00

Stunting is a condition of impaired growth and development in toddlers due to prolonged nutritional deprivation. In the Kota Timur Community Health Center, stunting is traditionally assessed based solely on body weight and height, neglecting other crucial factors such as socioeconomic status, maternal nutrition during pregnancy, history of illness, and dietary intake. This limited approach leads to inaccurate decision-making and misdiagnoses of stunting. This research applied the Multi-Attribute Utility Theory (MAUT) to identify stunting susceptibility levels in toddlers by integrating various determinants, including body weight, height, socioeconomic conditions, maternal nutrition during pregnancy, morbidity, and dietary intake. MAUT effectively integrates multiple criteria and manages data uncertainties through its utility concept, allowing for comparison across different alternatives to facilitate accurate decision-making. The results showed that Arbi, Manaf, and Aisyah were susceptible to stunting, with evaluation scores of 0.028, 0.288, and 0.299, respectively, while Daffa and Zayyan were not susceptible, with scores of 0.900 and 0.966, respectively. Therefore, the system utilizing MAUT to determine stunting susceptibility levels in toddlers can be adopted by health workers at the Kota Timur Community Health Center to enable efficient, quick, and accurate diagnosis by integrating multiple determinants of stunting susceptibility.

Social Media Analysis for Effective Information Dissemination and Promotions Using TOPSIS

2024-07-02T09:49:20+00:00

Social media has become an essential tool for spreading news and promotions. This research aims to evaluate the effectiveness of using social media as a strategy for disseminating information and promoting products using the TOPSIS method. Initial data was collected from a survey of social media users. The data was gathered through questionnaires distributed to various groups, including students, entrepreneurs, and office workers. The TOPSIS method was used to analyze the data and identify the most effective social media channels for information dissemination and promotion. The findings indicate that Facebook is the most effective platform for disseminating information, followed by Instagram and Twitter. Conversely, Instagram is the most effective platform for content promotion, followed by Facebook and YouTube. This study has significant implications for businesses and organizations that use social media for information dissemination and promotions. The TOPSIS (Technique for Order Preference by Similarity to Ideal Solution) method was used to evaluate and rank platforms based on criteria such as reach, accessibility, topicality, ease of use, creativity, informativeness, adaptability, transactionability, and security. The results show that TikTok is the best social media platform with the highest preference score of 0.755, followed by Facebook in second place, Instagram in third place, Twitter in fourth place, Telegram in fifth place, and YouTube in sixth place.

Comparative Study of Web Server Performance Testing with and without Docker Based on Virtual Machines

2024-07-02T10:25:11+00:00

Web server development is often hindered by the cost and resources required, as developing a web server typically requires a bare-metal server. Container technology, which allows for the development of multiple web servers on a single bare-metal server, has become popular. One of the most widely used containers is Docker. Docker reduces the need for costs and resources. Beyond the issues of cost and resource requirements, the performance of web servers also needs to be considered. The performance of web servers with and without Docker needs to be verified. This research aims to test the performance of two web servers, one using Docker and one not using Docker, utilizing the native hypervisor VMware ESXi. The web server performance test items in this study include CPU and RAM resource usage. The method for developing infrastructure systems uses SIDLC (System Infrastructure Development Life Cycle). Performance testing (Load Test) was conducted using Apache JMeter as a tool, with the manipulation of the number of threads predetermined. Resource usage information was monitored using Prometheus and Grafana. The research results show that with the same resources for each virtual machine, the CPU resource usage of Virtual Machine 2 (Undockerized) is less than that of Virtual Machine 1 (Dockerized). Meanwhile, RAM resource usage is not affected by the number of users on both virtual machines. Virtual Machine 2 (Undockerized) is better at handling HTTP requests. Virtual Machine 1 (Dockerized) can handle only 2,790 users, while Virtual Machine 2 (Undockerized) can handle more than 2,790 users without errors.