Ultimatics : Jurnal Teknik Informatika

Application Fuzzy AHP-TOPSİS Hybrid Method in Facility Location Selection for Software Systems

Kamala Rafig Aliyeva — Fri, 31 Jan 2025 00:00:00 +0000

Facility location is an integral part of the strategic planning process of almost every organization. Selecting the right location for software systems facilities involves considering various factors to ensure optimal performance, reliability, and cost-effectiveness. For business success, and competitive advantage there are some critical factors that very highly affect facility location. They are proximity to customers, infrastructure, labor quality, total cost, suppliers, etc. The criteria for selecting a facility location may be vaguely defined or open to interpretation. External factors such as economic conditions, political stability, and environmental risks may introduce vagueness and unpredictability into facility location decisions. These factors are often beyond the control of decision-makers but can have significant implications for the success of software systems. In this paper we apply fuzzy AHP-TOPSIS hybrid method for facility location in software systems. This offered model combines the fuzzy AHP and the fuzzy TOPSIS methods. Fuzzy AHP (Analytic Hierarchy Process) and fuzzy TOPSIS (Technique for Order of Preference by Similarity to Ideal Solution) are both decision-making methods commonly used in facility location selection. AHP helps in structuring a decision problem into a hierarchy of criteria and alternatives, while TOPSIS ranks alternatives based on their distance to the ideal solution. In the first part of the facility location selection process, we use fuzzy AHP method for determining weights of criteria that are important in selection process. Then by using fuzzy TOPSIS we rank alternatives and select appropriate location for facility.

Fuzzy TOPSIS Implementation for the Determination of Priority Scale in Improving Service Quality

Novrindah Alvi Hasanah, Muhammad Faisal, Linda Salma Angreani — Fri, 31 Jan 2025 00:00:00 +0000

Service quality plays a crucial role in economic development, particularly in the service industry, such as hotel services. Despite this, many hotels lack a systematic approach to help management identify areas that require improvement based on customer feedback. This research aims to develop a system that supports efforts to enhance service quality, utilizing the Fuzzy TOPSIS method. The study incorporates 150 data points obtained from questionnaires distributed to hotel service customers. The research involves two trials: service improvement priority and service eligibility. The results indicate an 84.45% accuracy level for service improvement priority testing, based on 120 out of 150 data points. Additionally, the accuracy level for service eligibility testing is 85.34%, derived from 131 data points out of the total 150. The research findings highlight the cafeteria as a significant area requiring improvement in service quality, aligning with the insights of hospitality experts. These results can serve as a foundation for management to enhance service quality based on selected criteria and alternatives.

Data Mining Klasifikasi Penjualan Motor Menggunakan Kombinasi Algoritma K-Means Dan Naïve Bayes

Eka Sofiati — Fri, 31 Jan 2025 00:00:00 +0000

The showroom, which has been established since 2006, which is located in Lapai, Padang city, has problems, namely the difficulty of analyzing consumer demand and a lot of accumulated sales data. In addition, there are many stocks of goods that are not available when consumer demand is high. From these problems a data mining application system is needed to improve sales patterns and process sales data to determine what is often purchased and not by using the data mining method, namely K-Means and Naïve Bayes. The data is obtained directly from CV. Unique Motor in the form of motorcycle sales data and motorcycle inventory data. At the system analysis stage, system design will be carried out using data mining using the K-Means and Naive Bayes algorithms. Where the program will be executed in the PHP and MySQL programming languages. The existence of a classification data mining system using a combination of K-Means and Naive Bayes can speed up the showroom in making decisions from the data taken so that the showroom can increase the number of stocks that have a hot-selling classification, so that the showroom not out of stock. A data mining system designed using a combination of KMeans and Naïve Bayes can assist showrooms in classifying motorcycle sales, as well as being able to align the availability and inventory of existing motorcycles by classifying sales volumes.

APPLICATION OF DEEP LEARNING TECHNIQUES FOR ENHANCING ARABIC VOCABULARY ACQUISITION IN STUDENTS AT MTS DARUN-NAJAH

Fri, 31 Jan 2025 00:00:00 +0000

Arabic vocabulary recognition is an important aspect of learning at MTs Darun - Najah, a school that emphasizes on Islamic religious education. This research proposes the application of Convolutional Neural Network (CNN) and EfficientNet B7 to create learning media for Arabic vocabulary recognition for students. This method is implemented in the form of a web-based application. The built application offers an innovative approach in learning by utilizing deep learning. The results of several trials conducted showed that the application of Convolutional Neural Network (CNN) and EfficientNet B7 achieved 90% accuracy with an average precision of 94.6%, recall 94.6%, and f1-score 94.6%. Tests using User Acceptence Testing (UAT) have a success accuracy rate of 87.2% which proves that users can accept quite well.

Sentiment Analysis in E-Commerce: Beauty Product Reviews

Gavrila Louise Tumanggor, Feliks Victor Parningotan Samosir — Fri, 31 Jan 2025 09:41:18 +0000

The increasing popularity of online shopping platforms is fueling the need for automated sentiment analysis for product reviews. This research aims to build an automatic sentiment analysis model in Indonesian for e-commerce product reviews. This model is expected to help consumers make purchasing decisions more quickly. We utilize the IndoBERT model, which has shown to be quite effective for general sentiment analysis, achieving an evaluation accuracy of 66.2% despite a high evaluation loss of 0.8006. The approach used combines Natural Language Processing (NLP) and Machine Learning (ML) techniques. It is hoped that this research will be useful for consumers, shop owners, and researchers in efficiently understanding the sentiment of e-commerce product reviews.

Prostate Cancer Screening for Specific Races Using Bioinformatics and Artificial Intelligence on Genomic Data

David Agustriawan — Fri, 31 Jan 2025 09:42:06 +0000

Prostate cancer is one of a deathly cancer worldwide. The higher incidence and mortality rate shows that it is an urgent call for all of us to fight against it in our own way. This study develops an artificial intelligence system to screening prostate cancer from normal patients in a specific race. Gene expression and its phenotype dataset was downloaded from xenabrowser.net Data preprocessing and filtering based on a particular race, bioinformatics computational analysis to determine the features and machine learning algorithm such as decision tree and random forest are used to develop AI model. All the procedure and analysis was performed using python programming The result show that only White and Black African American that has a proper number of dataset while Asian and American Indian has a very lack dataset. Differentially expression gene (DEG) analysis was performed to both White and Black African American cancer and normal dataset as a reference. 143 and 1 DEG are found in White and Black African American race respectively. ENSG00000225937.1 (PCA3) is identified as the highest up-regulated gene expression in cancer in both White and Black African American race. The results of DEG analysis then become features to develop Artificial Intelligence (AI) classification system. AI model was developed using decision tree and random forest with GriDSearch parameters optimization and stratified 10-fold cross validation. Both Decision tree and random forest model yield 96% accuracy in training dataset and 93% and 91% accuracy in testing dataset for decision tree and random forest, respectively.

Improved SVM for Website Phishing Detection Through Recursive Feature Elimination

Farica Perdana Putri, Feliciano Surya Marcello — Fri, 31 Jan 2025 09:42:26 +0000

Technology is developing faster every day, particularly in the information technology field. A website is one of the many information access points people use to do business activities, get information, and other purposes. Sophisticated websites are being developed and used, encouraging many naive individuals to commit crimes for financial gain. Phishing websites are a common method of using information technology to conduct fraud. One way to conduct phishing is by using the features on the website. One technique for identifying phishing websites is to use the Support Vector Machine (SVM) algorithm, which classifies websites based on features. However, the SVM algorithm is not able to detect many features so that the resulting accuracy and optimization level is also not good. Based on datasets, the SVM algorithm only gets around 60% to 70% accuracy. The use of Recursive Feature Elimination (RFE) feature selection is one way that can be done to cover the shortcomings of SVM. By eliminating features that irrelevant and redundance, RFE makes the SVM algorithm get a higher accuracy rate on the available dataset with an accuracy of 96.09%.

Implementation of Deep Learning Model for Identification of Skin Diseases by Utilizing Convolutional Neural Network

Lysa Apriani, Muhamad Bahrul Ulum — Fri, 31 Jan 2025 09:42:46 +0000

Skin diseases are health problems that affect many individuals worldwide. Rapid and accurate diagnosis of skin diseases is essential for effective treatment. In an effort to improve diagnosis, information technology and artificial intelligence have taken on increasingly significant roles. This study focuses on the implementation of deep learning models for skin disease identification using CNN architectures EfficientNetB0, Xception and VGG16. The models were trained and tested on a dataset of 1800 images with 5 dermatitis classes and 1 normal class. Confusion matrices were used to assess the performance of the three deep learning models on the components of accuracy, recall, precision, and F1-score. The results of the deep learning model that can classify dermatitis skin diseases with a performance of more than 90% for each evaluation matrix are deep learning models utilizing EfficientNetB0 transfer learning with an accuracy of 93%. In contrast, the Xception model indicates overfitting with a training accuracy of 99.96% and a validation accuracy of 86.38%. The VGG16 model indicates underfitting with a training accuracy of 69.71% and a validation accuracy of 46.79%.

Comparison of Multilinear Regression and AdaBoost Regression Algorithms in Predicting Corrosion Inhibition Efficiency Using Pyridazine Compounds

Yudha Mulyana, Muhamad Akrom, Gustina Alfa Trisnapradika, Nabila Putri Setiawan — Fri, 31 Jan 2025 09:43:10 +0000

Abstract-Corrosion is a serious problem in various industries that leads to increased production costs, maintenance, and decreased equipment efficiency. The use of organic compounds as corrosion inhibitors has become an increasingly desirable solution due to their effectiveness and environmental friendliness. This study compares the performance of two machine learning algorithms, Multilinear Regression (MLR) and AdaBoost Regression (ABR), in predicting the corrosion inhibition efficiency (CIE) of pyridazine-derived compounds. The dataset used consists of molecular properties as independent variables and CIE values as targets. To measure the performance of the model, a k-fold cross-validation process was used, where the dataset was divided into equal subsets. Each iteration uses one subset as validation data, while the other subset as training data. Results show that the AdaBoost Regression model achieves higher accuracy (99%) than Multilinear Regression (98%) in predicting CIE. Important feature analysis showed that Total Energy (TE) and Dipole Moment (µ) were the most influential variables in the ABR model, highlighting their important role in inhibitor effectiveness. Model evaluation was performed with R2 and RMSE metrics, where nonlinear models such as ABR were shown to be superior in predicting corrosion inhibition efficiency. These findings support the use of nonlinear methods to improve the effectiveness of protecting industrial equipment from corrosion.

Implementation of Gamification Method and Fisher-Yates Shuffle Algorithm for Design and Development Django Learning Application

Fri, 31 Jan 2025 00:00:00 +0000

The web framework emerges as a solution to enhance web development efficiency. Django, an open-source web framework written in the Python programming language, is one of the popular frameworks. Currently, there are not many programming learning platforms that provide specific programming learning materials for Django, implementing a method to boost user interest in using the platform. This research aims to design and build a web-based Django learning application using gamification methods designed based on the octalysis framework to enhance user learning interest. It also incorporates the Fisher-Yates shuffle algorithm to randomize questions for more variety. The application was tested by several users by filling out a questionnaire prepared using the Hedonic Motivation System Adoption Model (HMSAM). The evaluation results of the application obtained an average percentage of 84,15% in the aspect of behavioral intention to use, which means users strongly agree that the djangoing application generates a desire to use it again in the future. Furthermore, the results in the aspect of immersion were 81,44%, which means users agree that the djangoing application creates an immersive learning experience for the Django framework.

Leveraging Content-Based Filtering for Personalized Game Recommendations: A Flutter-Based Mobile Application Development

Alexander Waworuntu — Fri, 31 Jan 2025 09:44:19 +0000

The background of this study stems from the need for a recommendation system to assist users in finding games that match their interests. With the rapid growth of the gaming market, an increasing number of people engage in gaming activities. In 2022, the personal computer (PC) gaming market accounted for 37.9% of all gamers worldwide. One of the largest PC gaming platforms is Steam, developed by Valve Corporation, which boasts over 184 million active users. However, the overwhelming number of options can lead users to lose interest in purchasing games. Therefore, a recommendation system is required to help users find games that align with their preferences. The methods/theories employed in this study include data from the Steam Web API, SteamSpy API, and local JSON files. The Content-Based Filtering method, using the Cosine Similarity algorithm, was implemented to determine the similarity index between games and user preferences. Flutter was used for application development and to display the recommendation results to users. The results of this study show that the application was successfully developed, and the Content-Based Filtering method provided recommendations that met expectations. The highest cosine similarity factor achieved was 0.6454972244, indicating a fairly good level of accuracy. Application evaluation using the Technology Acceptance Model revealed positive reception, with a "Perceived Usefulness" score of 82.6% and a "Perceived Ease of Use" score of 86.2%, indicating that users found the application both useful and easy to use.