Garuda - Garba Rujukan Digital

Pemodelan Prediktif Emisi CO2 Kendaraan Kanada: Studi Komparatif Neural Network dan Support Vector Machine Rifki Nur Hidayat Putra; Nindya Dwi Lestari; Dinda Aprillia; Sumanto Sumanto; Imam Budiawan; Roida Pakpahan
IJAI (Indonesian Journal of Applied Informatics) Vol 10, No 1 (2025)
Publisher : Universitas Sebelas Maret

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.20961/ijai.v10i1.110736

Abstrak : Sektor transportasi merupakan penyumbang emisi karbon dioksida (CO2) terbesar yang memperparah perubahan iklim. Penelitian ini bertujuan mengembangkan model prediktif yang akurat untuk memperkirakan emisi CO2 kendaraan dengan memanfaatkan pendekatan pembelajaran mesin. Dataset yang digunakan adalah data emisi kendaraan Kanada dari Kaggle. Metode yang diterapkan adalah Support Vector Machine (SVM) dan Neural Network untuk menganalisis pola kompleks dari berbagai parameter teknis kendaraan, seperti ukuran mesin, jumlah silinder, dan jenis transmisi. Hasil penelitian menunjukkan bahwa Neural Network secara konsisten unggul dibandingkan SVM dengan tingkat akurasi prediksi melebihi 90% dan nilai F1-score mencapai 0,831 untuk model SVM serta 0,954 untuk model Neural Network, yang menunjukkan kinerja klasifikasi yang kuat dan konsisten. Neural Network juga terbukti lebih baik dalam menangkap hubungan non-linier antara karakteristik kendaraan dan emisi CO2. Keberhasilan model ini membuka peluang pengembangan model prediktif yang lebih canggih serta dapat menjadi dasar bagi pembuat kebijakan dalam merancang regulasi emisi yang lebih akurat dan berbasis data.=====================================================Abstract :The transportation sector is the largest contributor to carbon dioxide (CO2) emissions that exacerbate climate change. This research aims to develop an accurate predictive model to estimate vehicle CO2 emissions by utilizing a machine learning approach. The dataset used is Canadian vehicle emissions data from Kaggle. The methods applied are Support Vector Machine (SVM) and Neural Network to analyze complex patterns of various vehicle technical parameters, such as engine size, number of cylinders, and transmission type. The results showed that the Neural Network consistently excelled over SVM with a prediction accuracy rate exceeding 90% and an F1-score value of 0.831 for the SVM model and 0.954 for the Neural Network model, indicating a strong and consistent classification performance. Neural networks have also been shown to be better at capturing the non-linear relationship between vehicle characteristics and CO2 emissions. The success of this model opens up opportunities for the development of more sophisticated predictive models and can serve as a basis for policymakers to design more accurate and data-driven emissions regulations.

Solusi Virtual Try-On Kacamata Berbasis AI dengan Integrasi Model Deep Learning untuk E-Commerce Fashion Arnata Nur Rasyid; Asmawati Asmawati; Widya Viona Septi Tanjung; Sumanto Sumanto; Imam Budiawan; Roida Pakpahan
IJAI (Indonesian Journal of Applied Informatics) Vol 10, No 1 (2025)
Publisher : Universitas Sebelas Maret

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.20961/ijai.v10i1.110772

Abstrak : Banyak pengguna menghadapi kesulitan dalam memilih kacamata secara daring karena tidak dapat memastikan apakah model tertentu sesuai dengan bentuk wajah mereka. Masalah ini sering menimbulkan ketidakpuasan pelanggan dan tingginya tingkat pengembalian produk. Penelitian ini bertujuan untuk mengembangkan solusi Virtual Try-On kacamata berbasis kecerdasan buatan (AI), yang mengintegrasikan model deep learning untuk menciptakan pengalaman belanja daring yang lebih interaktif dan personal. Sistem bekerja dengan mendeteksi bentuk wajah dari foto yang diunggah pengguna menggunakan model Face Shape Detection yang telah dilatih dan mencapai akurasi hingga 89% kemudian memberikan rekomendasi kacamata yang paling cocok berdasarkan sistem rekomendasi Rule-Based. Setelah pengguna memilih salah satu produk dari daftar tersebut, sistem memanfaatkan AI Nano Banana untuk menggabungkan citra wajah dan produk kacamata secara realistis. Teknologi utama yang digunakan meliputi EfficientNetB2 sebagai model CNN utama, InsightFace untuk deteksi wajah presisi tinggi, dan AdamW sebagai algoritma optimasi. Hasil pengujian menunjukkan bahwa sistem ini efektif dalam menghasilkan visualisasi try-on yang akurat dan memuaskan, serta berpotensi meningkatkan konversi penjualan di platform e-commerce fashion.====================================================Abstract : Many users experience difficulty in selecting eyeglasses online due to the inability to determine whether a particular model suits their facial shape. This issue often results in customer dissatisfaction and high product return rates. This study aims to develop an AI-based virtual try-on solution for eyeglasses by integrating deep learning models to create a more interactive and personalized online shopping experience. The system functions by detecting the user’s face shape from an uploaded photo using a pre-trained Face Shape Detection model that achieves an accuracy of up to 89%, followed by a rule-based recommendation system that suggests the most suitable eyeglass frames. Once the user selects a product from the recommended list, the system utilizes AI Nano Banana to realistically generate a composite image of the user's face wearing the selected eyeglasses. The core technologies implemented include EfficientNetB2 as the primary CNN model for visual feature extraction, InsightFace for high-precision face detection, and AdamW as the optimization algorithm. Experimental results demonstrate that the system effectively generates accurate and realistic try-on visualizations, which are not only satisfactory to users but also have the potential to increase sales conversion rates in fashion e-commerce platforms.

Analisis Perbandingan Algoritma Random Forest, SVM, dan Logistic Regression untuk Menentukan Model Terbaik Prediksi Penyakit Diabetes Alghifar Firgiawan; Fauzan Nawwir Andriansyah; Raihan Naufal Ramadhan; Sumanto Sumanto; Imam Budiawan; Roida Pakpahan
Jurnal Teknik Informatika dan Teknologi Informasi Vol. 5 No. 3 (2025): Desember: Jurnal Teknik Informatika dan Teknologi Informasi
Publisher : Lembaga Pengembangan Kinerja Dosen

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.55606/jutiti.v5i3.6213

Diabetes is a chronic metabolic disorder characterized by elevated blood glucose levels caused by the body’s inability to produce or effectively respond to insulin. The increasing prevalence of diabetes in Indonesia requires accurate data-driven early detection systems to assist the diagnostic process. This study aims to compare the performance of three machine learning algorithms—Support Vector Machine (SVM), Random Forest, and Logistic Regression—in predicting diabetes disease based on patient clinical data. The dataset used was obtained from the Kaggle repository titled 100,000 Diabetes Clinical Dataset. The research process was conducted using the Orange Data Mining software through several stages, including data preprocessing, One-Hot Encoding transformation, model training, and evaluation using the 10-Fold Cross Validation method. The results show that the Random Forest algorithm achieved the best performance with an accuracy of 97.1%, followed by Logistic Regression at 96.0% and SVM at 92.3%. These findings indicate that ensemble-based methods such as Random Forest outperform others in producing stable and accurate predictions for diabetes diagnosis

Komparasi Algoritma Machine Learning (Random Forest, Gradient Boosting, dan Ada Boosting) untuk Prediksi Tingkat Penyakit Alzheimer Muhammad Raviansyah; Andika Amansyah; Farhan Fadhilah; Sumanto Sumanto; Imam Budiawan; Roida Pakpahan
Jurnal Teknik Informatika dan Teknologi Informasi Vol. 5 No. 3 (2025): Desember: Jurnal Teknik Informatika dan Teknologi Informasi
Publisher : Lembaga Pengembangan Kinerja Dosen

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.55606/jutiti.v5i3.6227

Alzheimer’s disease is one of the most common forms of progressive dementia and has become a major global health challenge as the aging population continues to increase. Early detection of this disease is crucial to mitigate its social, economic, and health impacts. In this context, data-driven approaches using machine learning algorithms can be utilized to predict Alzheimer’s risk more accurately. This study aims to compare the performance of three ensemble learning algorithms—Gradient Boosting, Random Forest, and AdaBoost—in predicting the risk level of Alzheimer’s disease using the public Alzheimer’s Disease Dataset, which includes demographic, clinical, and lifestyle data. The research process involved several stages, including data preprocessing, splitting data into training and testing sets, model training using cross-validation, and performance evaluation based on accuracy, precision, recall, F1-score, and AUC metrics. The experimental results show that the Gradient Boosting algorithm achieved the best performance with an accuracy of 0.956, an F1-score of 0.956, and an AUC of 0.985, demonstrating its ability to capture complex non-linear relationships among features such as age, MMSE score, and lifestyle factors. Meanwhile, Random Forest and AdaBoost achieved competitive yet slightly lower performance. These findings indicate that ensemble boosting approaches, particularly Gradient Boosting, hold great potential for medical decision-support systems in the early detection of Alzheimer’s disease and can serve as a foundation for developing more accurate and adaptive predictive models in the future.

Analisis Pola Pergerakan dan Prediksi Harga Emas Menggunakan Regresi Linear serta Model Time Series ARIMA dan VAR Roni Saputra Pratama; Ryehan Alfiansyah; Prasetyo Adi Suwignyo; Sumanto Sumanto; Imam Budiawan; Roida Pakpahan
Jurnal Teknik Informatika dan Teknologi Informasi Vol. 5 No. 3 (2025): Desember: Jurnal Teknik Informatika dan Teknologi Informasi
Publisher : Lembaga Pengembangan Kinerja Dosen

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.55606/jutiti.v5i3.6233

Gold is one of the most popular investment instruments due to its stable value and ability to protect assets against inflation. However, its price tends to fluctuate significantly, influenced by macroeconomic factors such as exchange rates, interest rates, and global geopolitical conditions. This study aims to analyze the movement patterns and predict gold prices based on historical data from 2019 to 2024 using the Linear Regression method and Time Series models, namely ARIMA and VAR. The analysis process was carried out using Orange Data Mining software, which enables the application of machine learning algorithms through a visual and interactive interface without manual coding. The dataset used consists of daily gold closing prices, processed and tested to evaluate model accuracy using Root Mean Square Error (RMSE) and Correlation Coefficient (R) indicators. The results indicate that the Linear Regression model effectively captures the general trend of gold prices, while ARIMA and VAR models produce more accurate forecasts based on historical fluctuations. The integration of regression and time series approaches improves prediction reliability. Overall, this research contributes to the development of financial data analysis and provides insights for investors in making more informed and data-driven investment decisions.

Penerapan Algoritma K-Means untuk Pengelompokan Kerentanan Wilayah terhadap Kasus DBD di Kota Bandung Zahwa Asfa Rabbani; Alya Avisa; Paulus Paulus; Sumanto Sumanto; Imam Budiawan; Roida Pakpahan
Jurnal Teknik Informatika dan Teknologi Informasi Vol. 5 No. 3 (2025): Desember: Jurnal Teknik Informatika dan Teknologi Informasi
Publisher : Lembaga Pengembangan Kinerja Dosen

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.55606/jutiti.v5i3.6239

Dengue Hemorrhagic Fever (DHF) is an infectious disease caused by the dengue virus and transmitted through bites of the Aedes aegypti mosquito. This illness remains a major public health concern in Indonesia, particularly in urban regions like Bandung City, where population density and environmental variations contribute to disease transmission. The purpose of this study is to apply the K-Means Clustering algorithm to group areas based on their level of vulnerability to DHF spread in Bandung City. The dataset, obtained from the Bandung Open Data portal covering the 2016–2024 period, was processed using the Orange Data Mining application. The analysis began with data preprocessing, which included cleaning, attribute selection, and normalization to ensure optimal clustering performance. The data were then grouped into three primary clusters representing high, medium, and low risk zones. The findings indicate that the K-Means algorithm effectively detects the spatial and temporal distribution of DHF cases and presents it through scatter plot visualizations that illustrate yearly patterns. High-risk regions are typically characterized by dense population, poor sanitation, and limited environmental management. These findings provide essential insight for local health authorities to design more targeted prevention and control strategies. Furthermore, this research can serve as a foundation for developing a decision support system that aids in monitoring, evaluating prevention efforts, and optimizing health resource allocation to reduce the incidence of DHF in the future.

Penerapan dan Perbandingan Algoritma SVM, Naive Bayes, dan Gradient Boosting dalam Prediksi Stroke Joseph Melchior Nababan; Iqro Mukti Arto; Putra Satria; Sumanto Sumanto; Imam Budiawan; Roida Pakpahan
Jurnal Teknik Informatika dan Teknologi Informasi Vol. 5 No. 3 (2025): Desember: Jurnal Teknik Informatika dan Teknologi Informasi
Publisher : Lembaga Pengembangan Kinerja Dosen

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.55606/jutiti.v5i3.6254

Stroke is a major cardiovascular disease that significantly contributes to global mortality and disability rates. Early detection through stroke risk prediction is essential in reducing its impact. This study focuses on evaluating and comparing the performance of three machine learning algorithms—Support Vector Machine (SVM), Naive Bayes (NB), and Gradient Boosting (GB)—in predicting stroke occurrence. The research utilizes the Healthcare Stroke Dataset, which contains 5,109 records and 11 predictor variables. Modeling was performed using Orange Data Mining software, with 70% of the data allocated for training and 30% for testing. The results show that the SVM algorithm achieved the highest performance, obtaining an AUC score of 0.919 and an accuracy of 96.0%, followed by Gradient Boosting with an AUC of 0.885 and accuracy of 95.2%, and Naive Bayes with an AUC of 0.803 and accuracy of 88.2%. Therefore, SVM is identified as the most effective algorithm for predicting stroke risk within this dataset.

Klastering Penyakit Diabetes Melitus dengan Algoritma K-Means berdasarkan Karakteristik Klinis Audy Aulia Azzahra; Fajar Yoga Adiansyah; Erlangga Rizki Ekaptra; Sumanto Sumanto; Imam Budiawan; Roida Pakpahan
Jurnal Teknik Informatika dan Teknologi Informasi Vol. 5 No. 3 (2025): Desember: Jurnal Teknik Informatika dan Teknologi Informasi
Publisher : Lembaga Pengembangan Kinerja Dosen

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.55606/jutiti.v5i3.6281

Diabetes Mellitus is a complex and progressive chronic metabolic disorder that requires a personalized management strategy tailored to each individual’s clinical, physiological, and lifestyle characteristics. Addressing this challenge, the present study aims to apply the K-Means algorithm to identify clustering patterns among diabetic patients using the Knowledge Discovery in Databases (KDD) framework. The dataset was obtained from the Kaggle repository, consisting of 769 patient medical records with key variables such as glucose levels, body mass index (BMI), blood pressure, age, and other metabolic parameters relevant to the diagnosis of Diabetes Mellitus. The research methodology includes several stages: data selection, preprocessing to handle missing values, duplication, and normalization to ensure the dataset is properly structured for analysis. The implementation of the K-Means algorithm was carried out using Orange Data Mining software to produce optimal clustering patterns. The analysis identified three primary clusters (C1, C2, C3) that demonstrated significant differences, particularly based on glucose levels as the dominant variable in cluster formation. The scatter plot visualization revealed clear separations among clusters, with high intra-cluster homogeneity and strong inter-cluster heterogeneity. These findings confirm the effectiveness of the K-Means algorithm as an unsupervised learning method capable of uncovering hidden patterns within clinical diabetes data. The results are expected to serve as a foundation for developing more adaptive and precise clinical decision support systems, assisting healthcare professionals in designing targeted management and intervention strategies aligned with each patient’s risk profile.

Penerapan Metode Logistic Regression untuk Memprediksi Potensi Penyakit Liver pada Pasien Tarmidzi Ibrahim; Imam Wahyudi; Vemi Januar Pratama; Sumanto Sumanto; Imam Budiawan; Roida Pakpahan
Jurnal Teknik Informatika dan Teknologi Informasi Vol. 5 No. 3 (2025): Desember: Jurnal Teknik Informatika dan Teknologi Informasi
Publisher : Lembaga Pengembangan Kinerja Dosen

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.55606/jutiti.v5i3.6284

Liver disease is a major global health concern that often goes undiagnosed in its early stages due to the absence of specific symptoms. Implementing data-driven approaches for early detection can significantly enhance diagnostic accuracy and improve clinical outcomes. This study aims to develop a predictive model using the Logistic Regression algorithm to identify individuals at high risk of liver disease. The data analysis process was conducted visually through data mining software, encompassing several stages such as data loading, feature selection, exploratory data analysis, and model evaluation. The dataset includes various clinical and laboratory attributes of patients, such as blood test results, liver function indicators, and demographic factors. The model’s performance was assessed using multiple evaluation metrics, with a focus on Classification Accuracy (CA) and the Area Under the ROC Curve (AUC) to measure predictive precision and classification ability. The results show that the Logistic Regression model achieved an accuracy of 71.8% and an AUC score of 0.746. These findings indicate that the model demonstrates good predictive performance and effectively identifies early-stage liver disease cases. However, further optimization is necessary to improve overall model efficiency and ensure more robust predictive capabilities in clinical applications.

Analisis Komparatif Sentimen Negatif Pengguna Platform E-Commerce Shopee dan Tokopedia selama Periode Diskon Faris Syahrendra; Cahyani Ayu Sulistyawati; Ginting Wibi Prasetyo; Sumanto Sumanto; Roida Pakpahan; Imam Budiawan
IJAI (Indonesian Journal of Applied Informatics) Vol 10, No 1 (2025)
Publisher : Universitas Sebelas Maret

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.20961/ijai.v10i1.110824

Abstrak : Fenomena potongan harga besar pada platform jual beli online sering kali menimbulkan kekecewaan bagi pengguna karena masalah dalam layanan, harga, dan pengiriman. Studi ini bertujuan untuk menganalisis dan membandingkan perasaan pengguna terhadap Shopee dan Tokopedia selama masa promosi dengan cara menggunakan pendekatan machine learning. Data ulasan diambil dari Google Play Store, yang terdiri dari 929 ulasan untuk Shopee dan 1.111 ulasan untuk Tokopedia. Dua algoritma untuk klasifikasi sentimen, yaitu Naive Bayes dan Neural Network, diimplementasikan dan dievaluasi dengan metode validasi silang 10-fold. Temuan yang berasal dari penilaian analitis menunjukkan bahwa model Naive Bayes menunjukkan tingkat akurasi dan presisi tertinggi yaitu 91,0%, sementara Neural Network memperoleh 83,9%. Selain itu, ulasan positif mendominasi sentimen terhadap Shopee (70%), sedangkan Tokopedia lebih banyak diwarnai oleh sentimen negatif (60%). Penemuan ini menandakan bahwa pengguna lebih puas dengan pengalaman diskon di Shopee dan memberikan masukan strategis untuk peningkatan layanan e-commerce.===============================================Abstract :Large-scale discount events on e-commerce platforms often lead to user disappointment due to issues with service, pricing, and delivery. This study aims to analyze and compare user sentiment towards Shopee and Tokopedia during promotional periods using a machine learning approach. Review data were sourced from the Google Play Store, consisting of 929 reviews for Shopee and 1,111 for Tokopedia. Two algorithms for sentiment classification, namely Naive Bayes and Neural Network, were implemented and evaluated using the 10-fold cross-validation method. Findings from analytical assessments indicate that the Naive Bayes model demonstrates the highest level of accuracy and precision at 91.0%, while the Neural Network obtained 83.9%. Furthermore, positive reviews dominated the sentiment towards Shopee (70%), whereas Tokopedia was largely characterized by negative sentiment (60%). These findings indicate that users are more satisfied with the discount experience on Shopee and provide strategic input for the improvement of e-commerce services.

Title

Found 11 Documents
Search

Abstract

Abstract

Abstract

Abstract

Abstract

Abstract

Abstract

Abstract

Abstract

Abstract

Title Search

Found 11 Documents Search

Abstract

Abstract

Abstract

Abstract

Abstract

Abstract

Abstract

Abstract

Abstract

Abstract

Title

Found 11 Documents
Search