Article Per Year (5 Year)
p-Index From 2021 - 2026
1.36
P-Index
This Author published in this journals
All Journal JOURNAL OF APPLIED INFORMATICS AND COMPUTING Jurnal Pengabdian pada Masyarakat Ilmu Pengetahuan dan Teknologi Terintegrasi Internet of Things and Artificial Intelligence Journal
Wardhana, Ajie Kusuma
Unknown Affiliation
Automotive Engineering Computer Science & IT Control & Systems Engineering Economics, Econometrics & Finance Education Electrical & Electronics Engineering Engineering

Published : 9 Documents Claim Missing Document
Claim Missing Document
Check
Articles

Found 7 Documents
Search
Journal : JOURNAL OF APPLIED INFORMATICS AND COMPUTING

 1 
Enhancing Eye Diseases Classification Using Imbalance Training & Machine Learning Ihwan, Muhammad Azrul; Wardhana, Ajie Kusuma
Journal of Applied Informatics and Computing Vol. 9 No. 4 (2025): August 2025
Publisher : Politeknik Negeri Batam

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.30871/jaic.v9i4.10207

Abstract

This research aims to evaluate the effectiveness of various machine learning algorithms in classifying eye diseases based on retinal images. The dataset comprises four categories of eye diseases: Cataract, Diabetic Retinopathy, Glaucoma, and Normal. The feature extraction method employed a transfer learning approach using ResNet50, followed by SMOTE for data balancing, PCA for dimensionality reduction, and normalization for scaling data consistently. Eleven machine learning models were evaluated, including basic algorithms, ensemble methods, and neural networks. The evaluation utilized metrics such as accuracy, precision, recall, and F1-score. K-Fold Cross Validation is also employed to observe all models' generalisation. The results revealed that the XGBoost algorithm achieved the highest performance with an accuracy of 92.03%, followed by LightGBM 91.88% and MLP 91.50%. K-Fold Validation also improved the MLP performance, which achieved an average accuracy of 91.94% with a standard deviation of 0.0178. This study successfully enhanced classification accuracy compared to previous studies and shows significant potential for clinical applications in resource-limited environments.
Performance Comparison of Machine Learning Algorithms Using EfficientNetB0 Feature Extraction on Dental Disease Classification Mustafa, Mohammad Faiq Ruliff; Wardhana, Ajie Kusuma
Journal of Applied Informatics and Computing Vol. 9 No. 5 (2025): October 2025
Publisher : Politeknik Negeri Batam

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.30871/jaic.v9i5.10286

Abstract

Oral health conditions such as dental caries, calculus, gingivitis, and ulcers are prevalent globally and require accurate early detection to prevent further complications. Traditional diagnostic methods such as visual inspection and manual radiograph analysis often rely on subjective judgment, leading to inconsistencies, delayed treatment, and limited accessibility, particularly in underserved areas. This study proposes an intelligent classification framework for dental disease detection based on intraoral images. Deep features were extracted using EfficientNetB0, followed by classification through eleven machine learning algorithms, including SVM, XGBoost, and K-Nearest Neighbors. Preprocessing steps included image augmentation, SMOTE for class balancing, and feature normalization. Among all models, SVM achieved the highest accuracy of 92,9%, while XGBoost and LightGBM followed closely at 91.3%. Using K-Fold Cross Validation, KNN algorithm has an increasing value with accuracy of 91,24%. This indicate the KNN algorithm able to tackle generalization problem towards the classification. The results demonstrate that features extracted using CNNs, when classified using machine learning algorithms, can provide a scalable and effective alternative to conventional diagnostic practices. Hence, Machine Learning algorithms provide a promising result towards dental disease classification.
Exploration of Machine Learning Algorithms and Class Imbalance Handling on Plant Disease Detection Aditya, Ervin; Wardhana, Ajie Kusuma
Journal of Applied Informatics and Computing Vol. 9 No. 5 (2025): October 2025
Publisher : Politeknik Negeri Batam

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.30871/jaic.v9i5.10338

Abstract

Plant leaf diseases pose a significant threat to agricultural productivity, necessitating accurate and efficient identification systems for timely intervention. This study proposes an approach that leverages deep feature extraction using a pretrained ResNet50 model combined with traditional machine learning algorithms to recognize 38 types of plant leaf diseases. Each image was transformed into a 2048-dimensional feature vector, followed by normalization and dimensionality reduction using Principal Component Analysis (PCA). To mitigate the issue of class imbalance in the dataset, random under-sampling was applied at the feature level to ensure equal representation across all classes. Eleven machine learning models were trained and evaluated using 5-fold cross-validation, with performance assessed through accuracy, precision, recall, F1-score, and ROC AUC score. Among the evaluated models, the Support Vector Machine (SVM) achieved the highest accuracy of 99.63%, followed by Logistic Regression at 97.33%, and LightGBM at 96.25%. These models demonstrated strong generalization capabilities in multiclass settings, while simpler classifiers like AdaBoost and Decision Tree yielded lower performance. A comparative analysis of training and test accuracy further highlighted model robustness and overfitting tendencies. The findings emphasize the potential of combining pretrained convolutional neural networks for feature extraction with conventional classifiers to address complex agricultural classification tasks. Future work may explore the inclusion of healthy leaf samples, alternative CNN architectures, and deployment in real-time diagnostic tools to support precision farming and improve crop health monitoring.
Image-Based Classification of Healthy and Unhealthy Goats Using ResNet-18 Deep Learning Model Amin, Nurrochim Amin Putra; Wardhana, Ajie Kusuma
Journal of Applied Informatics and Computing Vol. 9 No. 5 (2025): October 2025
Publisher : Politeknik Negeri Batam

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.30871/jaic.v9i5.10267

Abstract

Early detection of livestock health conditions is a critical factor in maintaining animal productivity and welfare. This study aims to develop an image-based classification system for identifying healthy and unhealthy goats using deep learning techniques. The dataset of goat images was obtained from Roboflow and processed through a series of augmentation, normalization, and feature extraction stages using the ResNet-18 convolutional neural network architecture pretrained on ImageNet. The dataset was divided into training and testing sets with a 70:30 stratified split to ensure balanced class distribution. To address class imbalance, a random undersampling technique was applied. The model was trained using optimally tuned hyperparameters, including the Adam optimizer, cross-entropy loss function, a batch size of 32, and 20 epochs. Evaluation results showed that the model achieved an accuracy of 95.97%, with a precision of 96.22%, recall of 95.97%, and F1-score of 95.92%. The confusion matrix and evaluation curves demonstrated that the model is both stable and reliable. These findings indicate that the proposed system has strong potential to be implemented in automated and real-time livestock health monitoring applications, providing a fast, accurate, and non-invasive solution for precision livestock farming.
Development of MobileNetV2 for CT-Scan Lung Classification Using Transfer Learning Rajendra, Daffa Fadhil; Wardhana, Ajie Kusuma
Journal of Applied Informatics and Computing Vol. 9 No. 5 (2025): October 2025
Publisher : Politeknik Negeri Batam

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.30871/jaic.v9i5.10282

Abstract

Lung cancer is one of the leading causes of cancer-related deaths worldwide, making early detection crucial for improving patient survival rates. This study proposes an automated classification approach based on deep learning using the MobileNetV2 architecture to identify three categories of lung CT scan images: normal, benign, and malignant. The dataset used is the augmented IQ-OTH/NCCD Lung Cancer Dataset, consisting of 3,609 images with a resolution of 224×224 pixels. All images underwent preprocessing steps including RGB conversion, pixel rescaling, and normalization. The MobileNetV2 model was modified by adding a GlobalAveragePooling2D layer, a dense layer, and dropout to reduce overfitting risk. Training was conducted for 28 epochs using the optimizer Adam, followed by evaluation using accuracy, precision, recall, and F1-score metrics. The model was tested on unseen data and validated using Stratified 5-Fold Cross Validation. The testing results showed an overall accuracy of 97%, with a perfect recall score (1.00) for the malignant class. The cross-validation yielded an average accuracy of 97.26% with a standard deviation of ±0.66%, indicating consistent model performance. Given its lightweight architecture and high accuracy, MobileNetV2 has the potential to be implemented as a decision support system in medical image analysis.
Addition of Non-Skin Classes in Skin Type Classification Using EfficientNet-B0 Architecture Sani, Haitsam Muftin; Wardhana, Ajie Kusuma
Journal of Applied Informatics and Computing Vol. 9 No. 5 (2025): October 2025
Publisher : Politeknik Negeri Batam

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.30871/jaic.v9i5.10335

Abstract

Skin type classification is an essential process in dermatology and skincare, aiming to categorize skin conditions such as dry, normal, and oily. However, image-based skin classification models often struggle when confronted with non-skin objects like clothing, background, or hair that are not accounted for in standard datasets. This study proposes a novel approach by integrating a nonskin class into a skin type classification model based on the EfficientNet-B0 architecture. The dataset used consists of images categorized into four classes: dry, normal, oily, and nonskin. The model was trained using transfer learning and optimized through techniques such as data augmentation, learning rate scheduling, and early stopping. The final evaluation achieved an accuracy of 91%, with the nonskin class showing perfect precision and recall. These results demonstrate that incorporating nonskin data can significantly enhance model robustness and accuracy. This research contributes a practical method for improving the reliability of skin classification systems, especially in real-world environments.
Hypertension Risk Prediction Using Stacking Ensemble of CatBoost, XGBoost, and LightGBM: A Machine Learning Approach Alfath, Abisakha Saif; Wardhana, Ajie Kusuma; Rumini, Rumini
Journal of Applied Informatics and Computing Vol. 9 No. 6 (2025): December 2025
Publisher : Politeknik Negeri Batam

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.30871/jaic.v9i6.10370

Abstract

Hypertension is a leading cause of cardiovascular diseases, chronic kidney failure, and strokes, affecting millions worldwide. Early detection and accurate risk prediction are crucial for effective management and prevention. This study aims to evaluate and compare the performance of different algorithms for predicting hypertension risk using a stacking ensemble approach. The model combines three gradient boosting algorithms XGBoost, LightGBM, and CatBoost as base learners, with Logistic Regression as the meta learner. The dataset, sourced from Kaggle, contains 4,240 instances with demographic and clinical attributes relevant to hypertension. The preprocessing steps included imputing missing values using the median, removing residual null entries, and addressing class imbalance through the SMOTE algorithm. Data were divided into 80% for training and 20% for testing. The evaluation showed that the stacking ensemble model achieved an overall accuracy of 92,65%, with precision, recall, and F1-scores consistently reaching 0.92 for both classes. The confusion matrix revealed minimal misclassification, indicating the model’s strong ability to differentiate between low and high risk individuals. These results emphasize that the primary goal of this research is to identify which algorithm provides the best performance for hypertension risk prediction. By evaluating and comparing different models, this study offers insights into choosing the most effective algorithm for clinical decision-making and early detection strategies.
Co-Authors Aditya, Ervin Alfath, Abisakha Saif Amin, Nurrochim Amin Putra Fikri, Muhammad Ainul Ihwan, Muhammad Azrul Mustafa, Mohammad Faiq Ruliff Ningrum, Fauzia Anis Sekar Rajendra, Daffa Fadhil Rauf, Budi Wijaya Riwanto, Yudha Rumini - Sani, Haitsam Muftin Yanuar Risca Pratiwi, Inggrid