Jurnal Transformatika
Not a member yet
319 research outputs found
Sort by
Identifikasi Penyakit Jantung Menggunakan Machine Learning: Studi Komparatif
No full textHeart disease is the number one cause of death globally. This condition is followed by an unhealthy lifestyle. Heart disease prediction needs to be done considering the importance of health. The presence of machine learning has made it easier for humans to make early detection of patterns approaching heart disease. This study compares 6 machine learning methods for disease classification with KNN, Naïve Bayes, Decision tree, Random forest, logistic regression, and SVM. The final classification obtained ranking accuracy with the highest value in the KNN method with precision, accuracy, re-call, fi-score tests. It is hoped that these results can be applied to real case studies of heart disease
Credit Risk Modeling Based on Geographical Location: A Case Study of Savings and Loan Cooperatives
Full text linkThe aim of this study is to examine how geographical location affects the credit risk faced by savings and loan cooperatives. Using a quantitative approach, this research will develop a credit risk model that considers geographical variables,measured by the Human Development Index (HDI). The initial stage of the research involves classifying the credit dataset according to the categoriesdetermined by Bank Indonesia. The data cleansing process resulted in attributes such as credit ceiling, HDI, and credit category. Analysis was conducted using Chi-Square, and Logistic Regression methods. The Chi-Square analysis results showed statistically significant relationship between credit ceiling, HDI, and credit category (p-value < 0.05). The Logistic Regression models demonstrated high accuracy in classifying the data, with Logistic Regression achieving 89.71%. In conclusion, credit ceiling and HDI have a significant influence on credit category, with the Logistic Regression model data classification. This study provides valuableinsights into how credit ceiling and HDI influence credit categories, which can be used to make better decisions related to public policy, developmentplanning, and social intervention
GA-SVM Wrapper Feature Selection untuk Penanganan Data Berdimensi Tinggi
Full text linkPeningkatan data dalam beberapa tahun terakhir ini mengalami peningkatan yang sangat signifikan karena penggunaan sosial media dan peralihan menjadi era digital. Teknik untuk pengolahan data menjadi informasi yang berguna dinamakan dengan data mining. Namun masalah yang terjadi ketika menerapkan data mining, khususnya metode klasifikasi adalah data berdimensi tinggi karena data berdimensi tinggi mempengaruhi hasil evaluasi dalam klasifikasi menjadi rendah. Data berdimensi tinggi didefinisikan sebagai data dengan jumlah fitur yang banyak dan kompleks, kompleksitas fitur mengakibatkan sulitnya memilih subset fitur yang optimal karena terdapat fitur yang tidak relevan. Dalam penelitian ini akan digunakan teknik wrapper dengan menerapkan metode metaheuristik yaitu algoritma genetika (GA) untuk pemilihan subset fitur agar lebih optimal, dan algoritma pengklasifikasi yang digunakan adalah algoritma Support Vector Machine (SVM), metode ini disebut dengan GA-SVM WFS. Hasil akurasi metode GA-SVM WFS lebih tinggi dibandingkan dengan metode SVM, dengan rata-rata hasil akurasi masing-masing sebesar 0,902 dan 0,874. Dalam penelitian ini terdapat perbedaan secara signfikan antara metode GA-SVM WFS dan metode SVM setelah dilakukan uji paired t-test dengan nilai p-value sebesar 0,01 dengan nilai α sebesar 0,05
GLOBAL THRESHOLDING IMPLEMENTATION FOR NOISE HANDLING IN DIGITAL IMAGE RECOGNITION
Full text linkText recognition (OCR - Optical Character Recognition) is a research field that is gaining widespread attention due to its wide application in image and document processing. Although OCR technology has achieved a high level of success, the main challenge faced is the presence of noise in text image, noise causes decreased text recognition results, noise causes miss classification. Therefore needed noise handling text recognition. The aim of this research is to provide valuable insight into the techniques and approaches used in the context of noise treatment using global threshold methods. The method used starts from an input digital image, then preprocessing is carried out by converting the image into a gray scale image, then a threshold is applied to the image, then recognition is carried out. From 6 experiments, the best results were obtained for character recognition with a threshold value (t) of 65 and a character recognition accuracy percentage of 94.29%. T value determined manually and static for separates the all object and the background, while in reality the lighting or contrast always varies. Suggestions for further research include developing an adaptive thresholding method approach to obtain threshold values automatically and optimally. So that if faced with varying lighting conditions or contrast, better results can be obtained
Pengembangan E-Modul Berbantuan Flipbook Berbasis Literasi Untuk Mata Kuliah Statistika
Full text linkStatistics is the basis for studying other subjects in the Informatics Engineering Study Program. However, some lecturers have not used teaching materials such as interactive modules to help students learn statistics. For this reason, research is needed which aims to (1) develop literacy-based Flipbook-assisted interactive e-modules in statistics courses that are valid, (2) determine the practicality of literacy-based Flipbook-assisted interactive e-modules in statistics courses for students. This type of research is Research and Development (R&D). The research subjects were 30 Informatics Engineering students taking Statistics courses. The development steps in this research are using ADDIE, namely Analysis, Design, Development, Implementation, and Evaluation. The research results obtained: 1) the literacy-based flipbook-assisted e-module that was developed was declared valid/feasible with an average percentage of material experts of 80% and media experts of 85.41%, 2) the flipbook-assisted interactive e-module that was developed met the criteria practically with a percentage reaching 75%
ANDROID BASED ADVERTISING REMINDER SYSTEM
Full text linkThe aim of this research is to introduce an Android-based reminder system that automatically notifies Surya Media Advertising employees of expired leases. Android is a "Linux-based operating system used on mobile devices such as smartphones and tablet computers (PDAs)"[8]. The method used in system development is the SDLC method which is the waterfall model. B. The design step must wait for the completion of the previous step, namely the requirements step. Visual Studio Code and Android Studio are used as software for this research [12]. The results of this study are; 1) Reminder system can help capture rentals for locations faster and easier; 2) The notification function makes it easy to know when the rental period has expired
IDENTIFIKASI PENYAKIT JANTUNG MENGGUNAKAN MACHINE LEARNING: STUDI KOMPARATIF
No full textHeart disease is the number one cause of death globally. This condition is followed by an unhealthy lifestyle. Heart disease prediction needs to be done considering the importance of health. The presence of machine learning has made it easier for humans to make early detection of patterns that are close to heart disease. Prediction of heart disease is important given the behavior of people who are still prone to risk factors. Conditions where predictions using machine learning for heart disease have not been compared with many using machine learning methods. Predictions of heart disease are needed along with the interrelationships of the variables. This research compares 6 machine learning methods for disease classification with KNN, Naïve Bayes, Decision tree, Random forest, logistic regression, and SVM. The final classification obtained ranking accuracy with the highest value of 82% in the KNN method with the confusion matrix test, precision, accuracy, re-call, and fi-score. These results can be applied to real case studies of heart diseas
Klasifikasi Jenis Buah Nanas Menggunakan Convolution Neural Network
Full text linkIndonesia is one of the countries with gread agricultural potential. One of the products of agriculture in Indonesia is pineapple. Pineapple is a tropical plant with edible fruit and one of the maximum economically vital plants in the Bromeliaceous family. The process of selecting pineapple species is generally very dependent on human perception. The development of technology and science makes it possible to perform classification or in terms of object selection using technology based on digital image-based characteristics. Images are used as a source of information that can be used to classify objects. One of the deep learning methods used is Convolutional Neural Networks (CNN) because they have a high deep network and are widely used to image data. Deep learning in Computer Vision has good capabilities in, one of which is image classification or object classification in images, and the network in CNN has a special layer, namely the convolution layer, The image convolution process in this study uses the keras package on GoogleColab, because making a neural network model using Keras does not need to write code to express mathematical calculations individually. Testing using a sample of 120 pineapple images shows an accuracy rate of 91,66% which is considered to be able to identify 3 types of pineapple fruit
Analisis Loyalitas Customer Perusahaan Konveksi dengan Model RFM dan Algoritma k-Means
No full textStrategi yang baik diperlukan suatu perusahaan dalam menjalankan usahanya. CV. Karunia Jaya merupakan salah satu perusahaan yang bergerak dalam bidang konveksi yanag menjual pakaian bayi. Dalam pelayanan terhadap customer CV. Karunia Jaya belum menerapkan strategi Customer Relationship Management (CRM). Untuk mengetahui loyalitas customer maka perlu dilakukan segmentasi pelanggan terhadap customer. Penelitian ini menggunakan data transaksi dari tahun 2021-2022. Algoritma k-means digunakan dalam penentuan cluster berdasarkan model Recency, Frequency, dan Moneetary (RFM), dibantu dengan tools Weka 3.8.6. Metode elbow digunakan untuk mencari jumlah cluster terbaik dari sekelompok data. Hasil dari penelitian ini yaitu terdapat 27 customer yang terbagi dalam tiga cluster, 21 customer potensi rendah, tiga customer potensi sedang, dan tiga customer potensi tinggi. Perusahaan dapat memberikan layanan yang berbeda terhadap setiap kelompok customer, sehingga hal tersebut dapat menguntungkan perusahaan
SISTEM MANAJEMEN PENEGAKAN DIAGNOSA PENYAKIT TYPUS DENGAN METODE NAIVE BAYES
Full text linkDalam dunia kesehatan, beberapa gejala yang timbul bisa menjadi tanda-tanda penyebab dari beberapa penyakit sekaligus. Kondisi yang dilematis ini sering kali mempersulit pengambilan keputusan mengenai penyakit yang diderita seseorang. Penentuan langkah pencegahan maupun pengobatan juga menjadi sulit untuk dilakukan, sehingga pemeriksaan lanjutan perlu dilakukan oleh tim medis. Sangat dimungkinkan jika tim medis kurang akurat dalam menentukan analisis penyakit dikarenakan munculnya gejala yang sama pada beberapa penyakit. Diperlukan sistem penegakan diagnose penyakit dengan metode na ve bayes akan mampu membantu pembelajaran bagi mahasiswa dalam mengenal gejala-gejala penyakit typus, dan juga dilengkapi dengan solusi pencegahan maupun solusi penanggulangan