Jurnal Sistem Cerdas

Classification of Hypertension Using Naïve Bayes Method with Data Discretization Approach Risk Factors

Yazid Munali, Armansyah — Mon, 29 Apr 2024 00:00:00 +0700

Generally, patients are unaware of their hypertension condition before having their blood pressure checked. One out of three Indonesians suffers from hypertension, and this figure continues to rise annually. Hypertension is often referred to as the silent killer because individuals with high blood pressure do not exhibit symptoms. This study aims to classify hypertensive patients in an effort to reduce the prevalence of hypertension in Indonesia by aiding in early detection of the disease and increasing awareness of hypertension among the Indonesian population. By using the Naïve Bayes method and implementing data discretization of risk factors, the dataset used comprises 11,627 health examination records of 4,434 participants from the Framingham Heart Study (FHS) organized by the National Institutes of Health. The classification method utilizes the Naïve Bayes Algorithm, and data discretization is performed using the CART (Classification and Regression Trees) method. The system provides an estimation of the probability of hypertension occurrence based on input factors/symptoms, where Naive Bayes achieves an accuracy rate 84.28%.

Study Program Selection Recommendation System Using the Fuzzy Inference System Mamdani

Abdullah Muadz Nadzir Azhar, Deden Pradeka, Devi Aprianti Rimadhani Agustini — Mon, 29 Apr 2024 00:00:00 +0700

This study presents a recommendation system for selecting study programs using the Mamdani fuzzy inference system. With the aim of assisting prospective students in making informed decisions, the system evaluates various factors including talents, UTBK scores, and school exam grades. The research utilizes the Temu Bakat test to assess talents and applies fuzzy logic to map inputs to outputs. Fuzzy rules are formulated based on the evaluation of antecedents, and aggregation combines multiple rules into a single output. Defuzzification converts fuzzy outputs into clear recommendations. The system's effectiveness is demonstrated through testing on students at UPI Campus in Cibiru, resulting in personalized study program recommendations for each student.

Design and Implementation of Continuous Settling Tank (CST) Temperature Monitoring System and Waste Pond Level in a Palm Oil Mills

Mon, 29 Apr 2024 00:00:00 +0700

Improving time efficiency and reducing work risks are crucial in palm oil mill operations. Two vital factors in palm oil mill operations are CST temperature and effluent pond level. Both of these factors have a significant impact on mill efficiency. Extreme temperatures, either too high or too low, can disrupt the settling and separation process of water, oil and sludge. In addition, overflowing effluent pond levels are often missed by technicians, so checking must be done manually. This research aims to develop a monitoring system that can monitor the temperature of the CST in safe conditions and measure the level of the waste pond in real-time by utilizing Internet of Things (IoT) technology. This control system relies on components such as the NodeMCU ESP8266 microcontroller, DS18B20 temperature sensor, and HC-SR04 proximity sensor. The method used is the design and manufacture of a prototype monitoring system. Field test results show that this system is able to make comparisons with manual measurement methods. In testing the CST temperature sensor and the waste pool level sensor, the accuracy rate reached 100%. The temperature was successfully monitored with a safe limit of 89°C, while the waste pool level was monitored through the monitoring website. This Internet of Things-based monitoring system has shown significant potential in improving the efficiency and safety of palm oil mill operations, and enables mill operators to effectively conduct remote real-time monitoring.

Implementation of the WASPAS Method in Selection Librarian FST UIN Sumatera Utara

Suendri, Anggika Wardani — Mon, 29 Apr 2024 00:00:00 +0700

The selection of prospective librarians is necessary to make the right decisions and determine the most suitable individual to meet the library's needs. However, in facing the complexity of changes in the information environment and technological developments, the FST library often needs help selecting prospective librarians using manual calculations, which can affect the effectiveness of the library head and staff in selecting. This research was conducted at the FST Library of the North Sumatra State Islamic University. This time, the researchers used the Weighted Aggregated Sum Product Assessment (WASPAS) method to select prospective librarians. The criteria used are communication skills (C1), mastery of technology (C2), customer service (C3), limited skills (C4), and lack of adaptability (C5). Then, the alternatives used were five samples of prospective librarians who registered at the FST library. The calculations using the WASPAS method showed that the alternative Nazla Maulida (A04) was a suitable candidate to become a librarian at the FST library with a total score of 0.854.

Design of Personalization Exam Classification Model Based on Imbalanced Class

Sri Anita, Arif Rachmat, Sunu Aditya Mahadany — Mon, 29 Apr 2024 00:00:00 +0700

Currently, personalized learning has become a necessity in the learning process. Research and implementation of personalization in the planning phases and implementation phases of learning have been extensively studied. However, that research has not yet reached the application stage in the learning assessment phase. Providing homogeneous examination material to each student has not considered the characteristics of learners. Even though the achievements from the assessment phase will provide a measure of the quality of the learning process as a whole. This research has analyzed the individual characteristics model, which is derived as a benchmark for identification of the information and characteristics of the test material, which is then formulated into a classification model based on supervised learning. This study identified text dataset questions and labeled unbalanced multi-classes. This presents a challenge to carry out experiments to find the most optimal data training strategy, the results provide optimal strategy combination results Logic: ENS, Verbal: ENS, Visual: CW+RES+ENS, Natural: CW+RES+ENS. Accuracy measurement results Logic (SVM): 0.85, Verbal (LR) 0.87, Visual (LR) : 0.93, Natural (NN) 0.93.

Classification of Beef and Pork with Deep Learning Approach

Akhiril Anwar Harahap, Rice Novita, Tengku Khairil Ahsyar, Zarnelly Zarnelly — Mon, 29 Apr 2024 15:17:15 +0700

Beef is one of the most consumed meats in Indonesia. However, the high price of beef has led to rogue traders mixing pork with beef. This condition occurs due to the lack of public knowledge about the difference between the two meats. To maintain food safety in Indonesia and especially in Riau province, the Livestock Service Office of Riau province conducts market surveys. There are several methods that are usually used to check the content of beef or pork, including Rapid Test Kit and Elisa. Both methods are time consuming and costly. One other solution that can be used is the artificial intelligence method, namely deep learning. In this research, a classification approach using deep learning is used to distinguish between beef and pork in the form of a web application. This research compares Convolutional Neural Network algorithm with Inception-V3 and Inception-Resnet-V2 architecture with hyperparameter optimization. From several experiments that have been carried out, the best model is the Inception-Resnet-V2 architecture with an experimental scenario using a learning rate of 0.001, and an optimizer Adam with an accuracy of 96.50%, Precision 96.48%, Recall 96.55% and F1-Score 96.50%. By using this model, web-based applications can be developed using the flask framework well and can perform classification accurately.

Sentiment Analysis ChatGPT Using the Multinominal Naïve Bayes Classifier (NBC) Algorithm

Dwi Sri Rahayu, Rice Novita, Tengku Khairil Ahsyar, Zarnelly — Mon, 29 Apr 2024 15:39:09 +0700

Chatbots have become one of the popular solutions for improving customer service. One well-known chatbot is ChatGPT, a language model developed by OpenAI. As time goes by and more and more people use ChatGPT, sentiment analysis is needed about users' opinions about the ChatGPT service. Therefore, it is necessary to carry out sentiment analysis of the ChatGPT service on Twitter to find out how users respond to this chatbot service. In this research, the results showed positive sentiment of 57%, negative sentiment of 29% and neutral sentiment of 14%. Topics for each sentiment were also obtained and sentiment prediction results from 40% of the test data with results of 96% positive, 3.5% negative and 0.5% neutral with a test accuracy of 63%.

A Library Locker Security System with Integrated RFID, Dual Camera Monitoring, and Telegram Notification

Mario Fazero Siregar, Jufrizel, Ahmad Faizal, Hilman Zarory — Mon, 29 Apr 2024 15:47:52 +0700

Abstract- Libraries generally provide lockers for users to store their luggage. Lockers in general still use conventional security by using manual locks. Currently, UIN Suska Riau Central Library still uses conventional key security which is considered easy to duplicate, lose, and damage. The purpose of this research is to improve locker security for user convenience and security through a system integrated with RFID, dual camera monitoring, and telegram notification. This tool has a CRUD (Create, Read, Update, Delete) system embedded in the telegram application managed by the admin. Admin can register, delete, and monitor locker user activities. This tool can identify registered and unregistered RFID tags properly. When a registered user accesses the locker, the system will send a picture of the user and a message to the telegram. Meanwhile, when an unregistered user accesses the locker 3 times incorrectly, the device will send the user's picture and message to the telegram and an alarm will sound for 3 seconds as an indication of an attempted locker break-in. The speed of information notification from the Internet of Things system has a delay in sending images with stability in the range of 3 - 4 seconds with relatively strong internet provider speed stability.

Keywords— locker security, internet of things, RFID, dual camera monitoring, telegram

An Implementation of Internet of Things for Digitalization of Kanban Production System

Victor Chris Samuel Purba, Virginia Lalujan — Tue, 30 Apr 2024 10:51:01 +0700

Just-In-Time production system in the form of kanban has been implemented in many production processes in order to optimize the production output. Internet of Things (IoT) has been integrated into production system in digitalizing production processes as well as obtaining and gathering real-time information in the factory floor through sensors, actuators, and network. In this study, an application of IoT into kanban and production system was proposed. The proposed electronic kanban system composed of Radio Frequency Identification (RFID) cards and readers, QR code and reader, microcontrollers, digital displays, and server. In this system, the kanban card is replaced by an RFID card. All workstations in the production process are equipped by RFID readers, microcontrollers, and displays to show relevant instructions to the operator. The microcontrollers are connected to a server to send information of the RFID card scanned by them and the time when the card was scanned. The latter information is used for tracing each of the products produced in each of the workstations. The proposed electronic Kanban system was implemented to an assembling process consisting of 7 workstations. The system was able to replace the conventional kanban card with RFID and electronic displays and performing the unit tracing capability.