Publication Search

Chronic diseases such as chronic kidney disease (CKD), diabetes, and heart disease remain major causes of mortality worldwide, highlighting the need for accurate and interpretable diagnostic models. However, conventional machine learning methods often face challenges of limited generalization, feature redundancy, and class imbalance in medical datasets. This study proposes an integrated classification framework that unifies three complementary feature paradigms: classical tabular attributes, deep latent features extracted through an unsupervised Long Short-Term Memory (LSTM) encoder, and quantum-inspired features derived from a five-qubit circuit implemented in PennyLane. These heterogeneous features are fused using a feature-wise attention mechanism combined with an AdaBoost classifier to dynamically weight feature contributions and enhance decision boundaries. Experiments were conducted on three benchmark medical datasets—CKD, early-stage diabetes, and heart disease—under both balanced and imbalanced configurations using stratified five-fold cross-validation. All preprocessing and feature extraction steps were carefully isolated within each fold to ensure fair evaluation. The proposed hybrid model consistently outperformed conventional and ensemble baselines, achieving peak accuracies of 99.75% (CKD), 96.73% (diabetes), and 91.40% (heart disease) with corresponding ROC AUCs up to 1.00. Ablation analyses confirmed that attention-based fusion substantially improved both accuracy and recall, particularly under imbalanced conditions, while SMOTE contributed minimally once feature-level optimization was applied. Overall, the attention-guided AdaBoost framework provides a robust and interpretable approach for clinical risk prediction, demonstrating that integrating diverse quantum, deep, and classical representations can significantly enhance feature discriminability and model reliability in structured medical data.

Breast cancer is the most prevalent cancer among women worldwide, requiring early and accurate diagnosis to reduce mortality. This study proposes a hybrid classification pipeline that integrates Hybrid Statistical Feature Selection (HSFS) with unsupervised LSTM-guided feature extraction for breast cancer detection using the Wisconsin Diagnostic Breast Cancer (WDBC) dataset. Initially, 20 features were selected using HSFS based on Mutual Information, Chi-square, and Pearson Correlation. To address class imbalance, the training set was balanced using the Synthetic Minority Over-sampling Technique (SMOTE). Subsequently, an LSTM encoder extracted non-linear latent features from the selected features. A fusion strategy was applied by concatenating the statistical and latent features, followed by re-selection of the top 30 features. The final classification was performed using a Support Vector Machine (SVM) with RBF kernel and evaluated using 5-fold cross-validation and a held-out test set. Experimental results showed that the proposed method achieved an average training accuracy of 98.13%, F1-score of 98.13%, and AUC-ROC of 99.55%. On the held-out test set, the model reached an accuracy of 99.30%, precision of 100%, and F1-score of 99.05%, with an AUC-ROC of 0.9973. The proposed pipeline demonstrates improved generalization and interpretability compared to existing methods such as LightGBM-PSO, DHH-GRU, and ensemble deep networks. These results highlight the effectiveness of combining statistical selection and LSTM-based latent feature encoding in a balanced classification framework.

Aspect-based sentiment Analysis (ABSA) is vital in capturing customer opinions on specific e-commerce products and service attributes. This study proposes a hybrid deep learning model integrating Bi-Directional Gated Recurrent Units (BiGRU) and Bi-Directional Attention Flow (BiDAF) to perform aspect-level sentiment classification. BiGRU captures sequential dependencies, while BiDAF enhances attention by focusing on sentiment-relevant segments. The model is trained on an Amazon review dataset with preprocessing steps, including emoji handling, slang normalization, and lemmatization. It achieves a peak training accuracy of 99.78% at epoch 138 with early stopping. The model delivers a strong performance on the Amazon test set across four key aspects: price, quality, service, and delivery, with F1 scores ranging from 0.90 to 0.92. The model was also evaluated on the SemEval 2014 ABSA dataset to assess generalizability. Results on the restaurant domain achieved an F1-score of 88.78% and 83.66% on the laptop domain, outperforming several state-of-the-art baselines. These findings confirm the effectiveness of the BiGRU-BiDAF architecture in modeling aspect-specific sentiment across diverse domains.

Monkeypox is a zoonotic disease caused by Orthopoxvirus, presenting clinical challenges due to its visual similarity to other dermatological conditions. Early and accurate detection is crucial to prevent further transmission, yet conventional diagnostic methods are often resource-intensive and time-consuming. This study proposes a deep learning-based classification model by integrating Xception and InceptionV3 using feature fusion to enhance performance in classifying Monkeypox skin lesions. Given the limited availability of annotated medical images, data augmentation was applied using Albumentation to improve model generalization. The proposed model was trained and evaluated on the Monkeypox Skin Lesion Dataset (MSLD), achieving 85.96% accuracy, 86.47% precision, 85.25% recall, 78.43% specificity, and an AUC score of 0.8931, outperforming existing methods. Notably, data augmentation significantly improved recall from 81.23% to 85.25%, demonstrating its effectiveness in enhancing sensitivity to positive cases. Ablation studies further validated that augmentation increased overall accuracy from 82.02% to 85.96%, emphasizing its role in improving model robustness. Comparative analysis with other models confirmed the superiority of our approach. This research enhances automated Monkeypox detection, offering a robust and efficient tool for low-resource clinical settings. The findings reinforce the potential of feature fusion and augmentation in improving deep learn-ing-based medical image classification, facilitating more reliable and accessible disease identification.

This study presents a Quantum Machine Learning (QML) architecture for perfectly classifying the Iris flower dataset. The research addresses improving classification accuracy using quantum models in machine-learning tasks. The objective is to demonstrate the effectiveness of QML approaches, specifically the Variational Quantum Circuit (VQC), Quantum Neural Network (QNN), and Quantum Support Vector Machine (QSVM), in achieving high performance on the Iris dataset. The proposed methods result in perfect classification, with all models attaining accuracy, precision, recall, and an F1-score of 1.00. The main finding is that the QML architecture successfully achieves flawless classification, contributing significantly to the field. These results underscore the potential of QML in solving complex classification problems and highlight its promise for future applications across various domains. The study concludes that QML techniques can offer transformative solutions in machine learning tasks, particularly those leveraging VQC, QNN, and QSVM.

Credit approval prediction is one of the critical challenges in the financial industry, where the accuracy and efficiency of credit decision-making can significantly affect business risk. This study proposes an outlier detection method using the Gaussian Mixture Model (GMM) combined with Extreme Gradient Boosting (XGBoost) to improve prediction accuracy. GMM is used to detect outliers with a probabilistic approach, allowing for finer-grained anomaly identification compared to distance- or density-based methods. Furthermore, the data cleaned through GMM is processed using XGBoost, a decision tree-based boosting algorithm that efficiently handles complex datasets. This study compares the performance of XGBoost with various outlier detection methods, such as LOF, CBLOF, DBSCAN, IF, and K-Means, as well as various other classification algorithms based on machine learning and deep learning. Experimental results show that the combination of GMM and XGBoost provides the best performance with an accuracy of 95.493%, a recall of 91.650%, and an AUC of 95.145%, outperforming other models in the context of credit approval prediction on an imbalanced dataset. The proposed method has been proven to reduce prediction errors and improve the model's reliability in detecting eligible credit applications.

Customer attrition has become the focus of many businesses today – since the online market space has continued to proffer customers, various choices and alternatives to goods, services, and products for their monies. Businesses must seek to improve value, meet customers' teething demands/needs, enhance their strategies toward customer retention, and better monetize. The study compares the effects of data resampling schemes on predicting customer churn for both Random Forest (RF) and XGBoost ensembles. Data resampling schemes used include: (a) default mode, (b) random-under-sampling RUS, (c) synthetic minority oversampling technique (SMOTE), and (d) SMOTE-edited nearest neighbor (SMOTEEN). Both tree-based ensembles were constructed and trained to assess how well they performed with the chi-square feature selection mode. The result shows that RF achieved F1 0.9898, Accuracy 0.9973, Precision 0.9457, and Recall 0.9698 for the default, RUS, SMOTE, and SMOTEEN resampling, respectively. Xgboost outperformed Random Forest with F1 0.9945, Accuracy 0.9984, Precision 0.9616, and Recall 0.9890 for the default, RUS, SMOTE, and SMOTEEN, respectively. Studies support that the use of SMOTEEN resampling outperforms other schemes; while, it attributed XGBoost enhanced performance to hyper-parameter tuning of its decision trees. Retention strategies of recency-frequency-monetization were used and have been found to curb churn and improve monetization policies that will place business managers ahead of the curve of churning by customers.

In the evolving landscape of agricultural technology, recognizing rice diseases through computational models is a critical challenge, predominantly addressed through Convolutional Neural Networks (CNN). However, the localized feature extraction of CNNs often falls short in complex scenarios, necessitating a shift towards models capable of global contextual understanding. Enter the Vision Transformer (ViT), a paradigm-shifting deep learning model that leverages a self-attention mechanism to transcend the limitations of CNNs by capturing image features in a comprehensive global context. This research embarks on an ambitious journey to refine and adapt the ViT Base(B) transfer learning model for the nuanced task of rice disease recognition. Through meticulous reconfiguration, layer augmentation, and hyperparameter tuning, the study tests the model's prowess across both balanced and imbalanced datasets, revealing its remarkable ability to outperform traditional CNN models, including VGG, MobileNet, and EfficientNet. The proposed ViT model not only achieved superior recall (0.9792), precision (0.9815), specificity (0.9938), f1-score (0.9791), and accuracy (0.9792) on challenging datasets but also established a new benchmark in rice disease recognition, underscoring its potential as a transformative tool in the agricultural domain. This work not only showcases the ViT model's superior performance and stability across diverse tasks and datasets but also illuminates its potential to revolutionize rice disease recognition, setting the stage for future explorations in agricultural AI applications.

This research aims to improve the effectiveness of lung cancer classification performance using Support Vector Machines (SVM) with hyperparameter tuning. Using Radial Basis Function (RBF) kernels in SVM helps deal with non-linear problems. At the same time, hyperparameter tuning is done through Random Grid Search to find the best combination of parameters. Where the best parameter settings are C = 10, Gamma = 10, Probability = True. Test results show that the tuned SVM improves accuracy, precision, specificity, and F1 score significantly. However, there was a slight decrease in recall, namely 0.02. Even though recall is one of the most important measuring tools in disease classification, especially in imbalanced datasets, specificity also plays a vital role in avoiding misidentifying negative cases. Without hyperparameter tuning, the specificity results are so poor that considering both becomes very important. Overall, the best performance obtained by the proposed method is 0.99 for accuracy, 1.00 for precision, 0.98 for recall, 0.99 for f1-score, and 1.00 for specificity. This research confirms the potential of tuned SVMs in addressing complex data classification challenges and offers important insights for medical diagnostic applications.

Rice plays a vital role as the main food source for almost half of the global population, contributing more than 21% of the total calories humans need. Production predictions are important for determining import-export policies. This research proposes the XGBoost method to predict rice harvests globally using FAO and World Bank datasets. Feature analysis, removal of duplicate data, and parameter tuning were carried out to support the performance of the XGBoost method. The results showed excellent performance based on which reached 0.99. Evaluation of model performance using metrics such as MSE, and MAE measured by k-fold validation show that XGBoost has a high ability to predict crop yields accurately compared to other regression methods such as Random Forest (RF), Gradient Boost (GB), Bagging Regressor (BR) and K-Nearest Neighbor (KNN). Apart from that, an ablation study was also carried out by comparing the performance of each model with various features and state-of-the-art. The results prove the superiority of the proposed XGBoost method. Where results are consistent, and performance is better, this model can effectively support agricultural sustainability, especially rice production.

Driving in a straight line is one of the fundamental tasks for autonomous vehicles, but it can become complex and challenging, especially when dealing with high-speed highways and dense traffic conditions. This research aims to explore the Deep-Q Networking (DQN) model, which is one of the reinforcement learning (RL) methods, in a highway environment. DQN was chosen due to its proficiency in handling complex data through integrated neural network approximations, making it capable of addressing high-complexity environments. DQN simulations were conducted across four scenarios, allowing the agent to operate at speeds ranging from 60 to nearly 100 km/h. The simulations featured a variable number of vehicles/obstacles, ranging from 20 to 80, and each simulation had a duration of 40 seconds within the Highway-Env simulator. Based on the test results, the DQN method exhibited excellent performance, achieving the highest reward value in the first scenario, 35.6117 out of a maximum of 40, and a success rate of 90.075%.

Music genre classification is one part of the music recommendation process, which is a challenging job. This research proposes the classification of music genres using Bidirectional Long Short-Term Memory (BiLSTM) and Mel-Frequency Cepstral Coefficients (MFCC) extraction features. This method was tested on the GTZAN and ISMIR2004 datasets, specifically on the IS-MIR2004 dataset, a duration cutting operation was carried out, which was only taken from seconds 31 to 60 so that it had the same duration as GTZAN, namely 30 seconds. Preprocessing operations by removing silent parts and stretching are also performed at the preprocessing stage to obtain normalized input. Based on the test results, the performance of the proposed method is able to produce accuracy on testing data of 93.10% for GTZAN and 93.69% for the ISMIR2004 dataset.

This research proposes a combination of Quantum Key Distribution (QKD) based on the BB84 protocol with Improved Logistic Map (ILM) to improve data transmission security. This method integrates quantum key formation from BB84 with ILM encryption. This combination creates an additional layer of security, where by default, the operation on BB84 is only XOR-substitution, with the addition of ILM creating a permutation operation on quantum keys. Experiments are measured with several quantum measurements such as Quantum Bit Error Rate (QBER), Polarization Error Rate (PER), Quantum Fidelity (QF), Eavesdropping Detection (ED), and Entanglement-based detection (EDB), as well as classical cryptographic analysis such as Bit Error Ratio (BER), Entropy, Histogram Analysis, and Normalized Pixel Change Rate (NPCR) and Unified Average Changing Intensity (UACI). As a result, the proposed method obtained satisfactory results, especially perfect QF and BER, and EBD, which reached 0.999.

Butterflies’ recognition serves a crucial role as an environmental indicator and a key factor in plant pollination. The automation of this recognition process, facilitated by Convolutional Neural Networks (CNNs), can expedite this task. Several pre-trained CNN models, such as VGG, ResNet, and Inception, have been widely used for this purpose. However, the scope of previous research has been somewhat constrained, focusing only on a maximum of 15 classes. This study proposes to modify the CNN InceptionV3 model and combine it with three data augmentations to recognize up to 100 butterfly species. To curb overfitting, this study employs a series of data augmentation techniques. In parallel, we refine the InceptionV3 model by reducing the number of layers and integrating four new layers. The test results demonstrate that our proposed model achieves an impressive accuracy of 99.43% for 15 classes with only 10 epochs, exceeding prior models by approximately 5%. When extended to 100 classes, the model maintains a high accuracy rate of 98.49% with 50 epochs. The proposed model surpasses the performance of standard pre-trained models, including VGG16, ResNet50, and InceptionV3, illustrating its potential for broader application.

This research introduces an image encryption scheme combining several permutations and substitution-based chaotic techniques, such as Arnold Chaotic Map, 2D-SLMM, 2D-LICM, and 1D-MLM. The proposed method is called Half-Inverted Cascading Chaos Cipheration (HIC3), designed to increase digital image security and confidentiality. The main problem solved is the image's degree of confusion and diffusion. Extensive testing included chi-square analysis, information entropy, NCPCR, UACI, adjacent pixel correlation, key sensitivity and space analysis, NIST randomness testing, robustness testing, and visual analysis. The results show that HIC3 effectively protects digital images from various attacks and maintains their integrity. Thus, this method successfully achieves its goal of increasing security in digital image encryption

Plant disease is one of the problems in the world of agriculture. Early identification of plant diseases can reduce the risk of loss, so automation is needed to speed up identification. This study proposes a custom-designed convolutional neural network (CNN) model for plant disease recognition. The proposed CNN model is not complex and lightweight, so it can be implemented in model applications. The proposed CNN model consists of 12 CNN layers, which consist of eight layers for feature extraction and four layers as classifiers. Based on the experimental results of a plant disease dataset consisting of 38 classes with a total of 87,867 image records. The proposed model can get high performance and not overfitting, with 97%, 98%, 97% and 97%, respectively, for accuracy, precision, recall and f1-score. The performance of the proposed model is also better than some popular pre-trained models, such as InceptionV3 and MobileNetV2. The proposed model can also work well when implemented in mobile applications.

Diabetes Mellitus is a hazardous disease, and according to the World Health Organization (WHO), diabetes will be one of the main causes of death by 2030. One of the most popular diabetes datasets is PIMA Indians, and this dataset has been widely tested on various machine learning (ML) methods, even deep learning (DL). But on average, ML methods are not able to produce good accuracy. The quality of the dataset and features is the most influential thing in this case, so deeper investment is needed to examine this dataset. This research will analyze and compare the PIMA Indians and Abelvikas datasets using the Random Forest (RF) method. The two datasets are imbalanced, in fact, the Abelvikas dataset is more imbalanced and has a larger number of classes so it is be more complex. The RF was chosen because it is one of the ML methods that has the best results on various diabetes datasets. Based on the test results, very contrasting results were obtained on the two datasets. Abelvikas had accuracy, precision, and recall, reaching 100%, and PIMA Indians only achieved 75% for accuracy, 87% for precision, and 80% for the best recall. Testing was done with 3, 5, 7, 10, and 15 tree number parameters. Apart from that, it was also tested with k-fold validation to get valid results. This determines that the features in the Abelvikas dataset are much better because more complete glucose features support them.

Skin is the largest organ in humans, it functions as the outermost protector of the organs inside. Therefore, the skin is often attacked by various diseases, especially cancer. Skin cancer is divided into two, namely benign and malignant. Malignant has the potential to spread and increase the risk of death. Skin cancer detection traditionally involves time-consuming laboratory tests to determine malignancy or benignity. Therefore, there is a demand for computer-assisted diagnosis through image analysis to expedite disease identification and classification. This study proposes to use the K-nearest neighbor (KNN) classifier and Gray Level Co-occurrence Matrix (GLCM) to classify these two types of skin cancer. Apart from that, the average filter is also used for preprocessing. The analysis was carried out comprehensively by carrying out 480 experiments on the ISIC dataset. Dataset variations were also carried out using random sampling techniques to test on smaller datasets, where experiments were carried out on 3297, 1649, 825, and 210 images. Several KNN parameters, namely the number of neighbors (k)=1 and distance (d)=1 to 3 were tested at angles 0, 45, 90, and 135. Maximum accuracy results were 79.24%, 79.39%, 83.63%, and 100% for respectively 3297, 1649, 825, and 210. These findings show that the KNN method is more effective in working on smaller datasets, besides that the use of the average filter also has a significant contribution in increasing the accuracy.

Convolutional neural network (CNN) is a deep learning (DL) model that has significantly contributed to medical systems because it is very useful in digital image processing. However, CNN has several limitations, such as being prone to overfitting, not being properly trained if there is data duplication, and can cause unwanted results if there is an imbalance in the amount of data in each class. Data augmentation techniques are used to overcome overfitting, eliminate data duplication, and random under sampling methods to balance the amount of data in each class, to overcome these problems. In addition, if the CNN model is not designed properly, the computation is less efficient. Research has proved that data augmentation can prevent or overcome overfitting, eliminating duplicate data can make the model more stable, and balancing the amount of data makes the model unbiased and easy to learn new data as evidenced through model evaluation and testing. The results also show that the custom convolutional neural network model is the best model compared to ResNet50 and VGG19 in terms of accuracy, precision, recall, F1-score, loss performance, and computation time efficiency