Ultimatics : Jurnal Teknik Informatika

A Comparative Study of Body Motion Recognition Methods for Elderly Fall Detection: A Review

2024-07-16T02:43:28+00:00

To maintain the welfare of the elderly, intensive and effective monitoring is needed to ensure their safety. Conventional elderly activity monitoring has several limitations (i.e., space and time) due to human abilities. This problem can be overcome by applying real-time monitoring methods using Wireless Body Area Networks (WBAN) and Artificial Intelligence (AI). Several methods have been used and tested, including artificial intelligence implementations from sensor data-based to computer vision-based pattern recognition for body motion classification. Several methods that have been studied show accurate results in classifying elderly body motions/gestures. However, the Human Activity Recognition (HAR) method performs better for elderly activity monitoring applications and makes fall classification more accurate.

Application of Convolutional Neural Network Using TensorFlow as a Learning Medium for Spice Classification

2024-07-16T02:43:28+00:00

The purpose of this research are: (1) To determine the accuracy of the CNN method in the development of a website for classifying spices, (2) To assess the feasibility of the spice classification website as a learning medium, (3) To ascertain user responses to the spice classification website as a learning medium. The method employed in this research is research and development. This study utilizes the ADDIE development method, which comprises 5 stages: (1) Analysis, (2) Design, (3) Development, (4) Implementation, and (5) Evaluation. The research yielded a significantly high accuracy rate. This is demonstrated by the results showing an accuracy of 96%, precision of 97%, and recall of 96%. Moreover, the research found the developed website to be feasible. This is supported by the evaluation using the Learning Object Review Instrument (LORI), resulting in a score of 88% from media experts and a score of 90% from subject matter experts. Additionally, user response was positive. This is evidenced by testing the learning media on 10th-grade culinary students from SMK N 4 Surakarta, which yielded a score of 76% using the System Usability Scale (SUS), indicating a favorable usability assessment. In conclusion, the spice classification website, as a learning medium, can be employed as a suitable educational tool.

Comparing Karate Framework with Others for Automated Regression Testing: A Case Study of PT Fliptech Lentera Inspirasi Pertiwi

2024-07-16T02:43:28+00:00

In the rapidly evolving digital era, applications, and software systems increasingly rely on Application Programming Interfaces (APIs) to enable interaction, integration, and functionality extension. However, manual testing of APIs is often inefficient and challenging to reuse when changes occur. To address this, automation testing has become a more effective choice, where test scripts can verify and execute tests repeatedly, easily adapting to API changes. Essentially, automation testing plays a vital role in software maintenance, particularly in regression testing, which tests modified or upgraded software versions to ensure that their core functions remain unchanged and unaffected. One approach to automation testing is employing the Software Testing Life Cycle (STLC), which follows a systematic series of stages conducted by the testing team to ensure software product quality. This paper utilizes PT Fliptech Lentera Inspirasi Pertiwi’s public API to conduct testing on 25 scenarios from two modules. The objective is to utilize the Karate Framework to conduct these automated regression tests, resulting in an impressively short testing duration, averaging only 42.645 seconds, or approximately 1.706 seconds per scenario. A comparison with the Behave framework, using the same scenarios but with differences in steps, reveals that Behave achieves a duration of 18.762 seconds, or 0.750 seconds per scenario, making it 127.295% faster than Karate. However, in terms of the number of steps, Behave covers only 188, while Karate includes 543. This means that Behave requires 0.100 seconds per step, while Karate necessitates 0.079 seconds per occurrence. Karate provides more detailed results by 188.830% per step or 26.582% in terms of step duration. The primary goal is to enhance testing efficiency, expedite issue identification and resolution, provide a clearer testing process, and potentially improve overall software quality.

Educational Game Design For Carbon Emission Using Game Development Life Cycle Method

2024-07-16T02:43:28+00:00

Carbon emissions are gases that arise from human actions, such as burning fossil fuels and industrial waste. Climate change is currently a problem that is increasingly attracting the attention of the wider community worldwide, including Indonesia. The educational game about carbon emissions applies the Game Development Life Cycle (GDLC) approach, which consists of six stages, including initiation, pre-production, production, testing, beta, and launch. The educational game on carbon emissions is expected to help raise youth awareness about the importance of reducing carbon emissions and provide information about efforts to reduce carbon emissions for the younger generation and the general public.

Designing a QR Code Attendance System Using BYOD (Bring Your Own Device)

2024-07-16T02:43:28+00:00

Attendance is an activity of collecting attendance data from each individual who attends events, work, and learning. The current application of attendance in certain companies, schools, or universities is still done manually using paper so it is considered less efficient and effective. Digitizing attendance activities can provide many benefits, such as making managing large amounts of attendance data easier. This is usually used in companies or schools. To reduce additional costs, this can be done by using a personal device as a medium for taking attendance, this can be called BYOD or Bring Your Own Device. The attendance that will be designed will use the user's smartphone or mobile device as a medium for taking attendance by scanning the QR code. The results of tests carried out using black box testing on mobile and web applications, shows that all the features contained in both applications are running according to their function. The use of QR Codes and also the implementation of BYOD can make it easier for users to take attendance. Apart from this, it is also easier for admins to manage user attendance data.

U-TAPIS Sal-Tik : Typing Error Detection Using Random Forest Algorithm

2024-07-16T02:43:28+00:00

The development of technology in the field of journalism has grown very rapidly. However, in the field of journalism there are still frequent deviations from the language on online news portals. This can be seen from the aspect of spelling and word usage. Spelling mistakes that occur in the news can cause the information contained in the news to be unclear and ambiguous. Based on these problems, a study was conducted to create a model to detect type error in Indonesian. This model is created using the random forest algorithm. random forest is an algorithm that works by building several decision trees and then combining the decisions from each tree that has been built and taking the most votes from the predictions of each tree so that it will produce stable and accurate predictions. The results of the accuracy of the model in the research that has been done is 100%. However, it should be noted that this 100% result is that the model is able to detect words that are already contained in the dataset. Based on the evaluation results that have been carried out, because the detected word is contained in the dataset, the accuracy issued is 100%. The built model successfully detects type error in Tribunnews news articles.

Recommendation System Coffee Shop using AHP and TOPSIS Methods

2024-07-16T02:43:29+00:00

Indonesian people generally like to spend time with friends, family and business colleagues while drinking coffee. This habit of consuming coffee can not only be done at home, but can also be done in other places such as traditional and modern coffee shops. This has also significantly influenced the growth of coffee shops, especially in Tangerang. So people are faced with so many choices and alternative coffee shops to visit. This research was conducted to create a system that can recommend coffee shops in Tangerang based on priority criteria input by the user. Therefore, this recommendation system uses the Multi Criteria Decision Making (MCDM) method, where the process of making decisions is based on several criteria. This research uses the method Analytical Hierarchy Process (AHP) and Technique for Order of Preference by Similarity to Ideal Solution (TOPSIS). This research was tested using the Usefulness, Satisfaction, and Ease of Use (USE) Questionnaire and received a very good rating with an overall score of 87.6\%, so the conclusion was that the average respondent felt helped by this recommendation system.

Sentiment Analysis of IMDB Movie Reviews Using Recurrent Neural Network Algorithm

2024-07-16T02:43:29+00:00

IMDb is a well-known platform that provides user reviews and ratings of various movies. The number of reviews found on IMDb is quite large, reaching thousands of reviews. Although a movie can have a high overall rating, it is still possible to receive negative reviews from some viewers. Therefore, the purpose of this sentiment classification system is to provide a benchmark for the level of sentiment contained in the movie, and hope that filmmakers can use this information as a reference in the development of their next movie. In this research, reviews from IMDb users are classified into two types, namely positive reviews and negative reviews. The program was created using the Python language with the LSTM (Long Short-Term Memory) classification model of the RNN (Recurrent Neural Network) algorithm. The purpose of using this algorithm is to measure the level of prediction accuracy in the classification process. The results of three test ratios, namely 60:40, 70:30, and 80:20, show that in the scenario of 80% data training and 20% data testing has better performance with the results accuracy of 96%, precision of 97%, recall of 98%, f1-score of 97%.

Comparison of Fine-tuned CNN Architectures for COVID-19 Infection Diagnosis

2024-07-29T03:47:05+00:00

SARS-CoV-2 (COVID-19) virus spread quickly worldwide affects a variety of industries. The government took preventive steps to control the infection, such as diagnosing the human's lung by taking an X-Ray to see if the lungs were infected with COVID-19 or not. Using several pre-trained Convolutional Neural Network models as the basic model, this research deconstructs the comparison of fine-tuned architecture to identify which pre-trained model delivers the best outcomes in diagnosis by applying machine learning. Comparison is conducted using two scenarios that use batch sizes 64 and 32. Accuracy and f1 score are two evaluation metrics used to justify the model's good performance because the images in the real world, especially for positive classes, are scarce. According to the study, EfficientNetB0 outperforms other pre-trained models, namely ResNet50V2 and Xception, which achieved an accuracy of 0.895 and f1 score of 0.8871.

Public Sentiment Analysis on the Transition from Analog to Digital Television Using the Random Forest Classifier Algorithm

2024-07-16T02:43:29+00:00

Television is one of the most popular media for entertainment and information. Analog television is the most widely used type among the public. However, with technological advancements, analog television is becoming obsolete and is being replaced by digital television, which offers better performance. On November 2, 2022, the Government officially mandated the transition from analog to digital broadcasting. This Analog Switch Off program has elicited various pro and con opinions among the public. Twitter, a widely used social media platform, facilitates rapid communication and information dissemination among users. This study aims to classify public sentiment regarding the Analog Switch Off policy as either positive or negative. The classification model used is the Random Forest algorithm, with the Lexicon Inset for data labeling, Count Vectorizer and TF-IDF Vectorizer for data vectorization and weighting, and various train-test data splits. The study achieved the best classification performance using the Count Vectorizer method, with an 80%:20% train-test data ratio, yielding an accuracy of 88%, precision of 88%, recall of 88%, and an F1-score of 88%.

Index Terms—Analog Television; Digital Television; Sentiment; Twitter; Random Forest