IAES International Journal of Artificial Intelligence (IJ-AI)
Not a member yet
1769 research outputs found
Sort by
Integrating random forest and genetic algorithms for improved kidney disease prediction
Get PDFThis work offers a novel method for predicting chronic kidney disease (CKD) by combining random forest (RF) classification with genetic algorithm (GA) to optimize important parameters. The dataset comprises 1,659 patients with 51 clinical parameters. The suggested method emphasizes the optimization of random state values, test size, and essential hyperparameters, such as the number of trees in the forest, the least number of samples needed at a leaf node, and the smallest number of samples necessary to split an internal node. The optimization process is conducted in two stages: the first stage optimizes the random state and test size, while the second stage focuses on hyperparameters. Through extensive simulations over 50 runs, the study demonstrates that the optimized model achieves an accuracy ranging from 0.9451 to 0.9738. The results indicate a maximum increase in accuracy of 2.09%, showcasing the effectiveness of the GA-RF integrated approach in enhancing model performance. This work provides valuable insights into the impact of parameter optimization on machine learning (ML) models, particularly in medical diagnostics, and offers a robust framework for developing highly accurate predictive models
Unveiling critical features for failure prediction in green internet of things applications
Get PDFThe rapid growth of the green internet of things (GIoT) in recent years signifies a transformative shift in internet of things (IoT) solution development. This evolution is driven by technological advancements, heightened environmental awareness, and a global imperative to combat climate change. Ensuring the reliability of GIoT applications is crucial for their success. This study identifies critical features for predicting IoT device failures, enabling early detection and intervention. Using datasets from industry, energy, and agriculture sectors, we employ a feature selection strategy to analyze extensive data from diverse GIoT deployments. Our analysis identifies significant features and integrates key insights from existing literature. Our findings support enhanced predictive maintenance strategies, reduced downtime, and improved overall performance of sustainable IoT solutions
Prediction of side effects of drug resistant tuberculosis drugs using multi-label random forest
Get PDFDrug-resistant tuberculosis (DR-TB) has become a concern because anti-tuberculosis drugs (ATD) used to treat it can cause side effects in patients. This study aimed to predict the potential side effects of ATD using a multi-label classification approach with a random forest (RF) algorithm. This study used 660 medical record data, including the 14 ATD treatments prescribed to the patients and the six side effects experienced by patients. The model was trained using the best parameters based on the hyperparameter tuning process. The results show that the RF multi-label algorithm can be an alternative for building ATD side effect prediction models because it produces the most optimal performance value compared to the decision tree (DT) and extreme gradient boosting (XGBoost). The area under the curve (AUC) score of all RF multi-label models is above 0.8, which means that all RF multi-label models are considered acceptable and applicable for ATD side effect prediction. In addition, eight features influenced the models based on the average feature importance score of the RF models. This study is expected to help predict the side effects of ATD used to treat DR-TB based on ATD treatment and determine the most promising tree-based machine learning algorithm for predicting ATD side effects
Fuzzy risk assessment system for indoor air quality and respiratory disease prevention
Get PDFThis study addresses the evaluation of indoor air quality, with a focus on mitigating respiratory diseases and sick building syndrome (SBS). Recognizing that different pollutants exhibit variable behavior depending on environmental factors and human activity, the objective was to develop a fuzzy logic-based classification system that integrates environmental variables such as temperature, relative humidity, and pollutant concentrations‒particulate matter (PM10, PM2.5), carbon dioxide (CO₂), and total volatile organic compound (TVOC)‒into a unified model. The method involved defining risk levels as low, moderate, high, and very high, and implementing 54 fuzzy rules to dynamically and accurately categorize these risks, based on measurements taken between 2022 and 2024 in the states of Morelos and Puebla under various relative humidity and temperature scenarios. The analysis of the results demonstrated robust system performance, with an overall accuracy of 94.08%, but also revealed challenges in distinguishing between adjacent risk classes. This research contributes to a better understanding of the complex impacts of air quality on health and reinforces efforts to mitigate respiratory problems and SBS in densely populated indoor environments
Using the ResNet-50 pre-trained model to improve the classification output of a non-image kidney stone dataset
Get PDFKidney stone detection based on urine samples seems to be a cost-effective way of detecting the formation of stones. Urine features are usually collected from patients to determine if there is a likelihood of kidney stone formation. There are existing machine learning models that can be used to classify if a stone exists in the kidney, such as the support vector machine (SVM) and deep learning (DL) models. We propose a DL network that works with a pre-trained (ResNet-50) model, making non-image urine features work with an image-based pre-trained model (ResNet-50). Six urine features collected from patients are projected onto 172,800 neurons. This output is then reshaped into a 240 by 240 by 3 tensors. The reshaped output serves as the input to the ResNet-50. The output of this is then sent into a binary classifier to determine if a kidney stone exists or not. The proposed model is benchmarked against the SVM, XGBoost, and two variants of DL networks, and it shows improved performance using the AUC-ROC, Accuracy and F1-score metrics. We demonstrate that combining non-image urine features with an image-based pre-trained model improves classification outcomes, highlighting the potential of integrating heterogeneous data sources for enhanced predictive accuracy
Temporal context of lightweight network model for detecting boats approaching the tsunami early warning system
Get PDFThe tsunami early warning system (TEWS) is a device that detects potential tsunamis. However, a boat that approaches TEWS is a source of communication disturbance. A convolutional neural network (CNN), as part of intelligent computer vision, is one solution for detecting boats and providing a warning to move away from the TEWS area. Water segmentation and refinement-temporal (WaSR-T), as the current advanced CNN network, exhibits impressive performance in detecting object obstacles in the marine domain, although it requires a powerful computational device. In the paper, we propose a modification of WaSR-T, replacing the most computationally intensive stages with a lightweight version called lightweight WaSR-T. On the proposed lightweight WaSR-T, the previous encoder of WaSR-T was replaced with MobileNetV3, and some feature layer maps were reduced as input to the decoder. For training and validating the lightweight WaSR-T, the image dataset representing the open sea and our extended dataset from Indonesia's ocean region were used. Based on the quantitative results and evaluation of the computational load, the sensitivity to detect a boat for WaSR-T and lightweight WaSR-T is 95.71% and 90.00%, respectively. The lightweight WaSR-T required less memory at 32.57%, resulting in a 0.0761% reduction in total processing time compared to the original WaSR-T. Therefore, our proposed lightweight WaSR-T is promising for use as the central part of an intelligent maritime computer vision system in TEWS
Data-driven support vector regression-genetic algorithm model for predicting the diphtheria distribution
Get PDFIndonesia is one of the countries with the largest number of diphtheria sufferers in the world. Diphtheria is a case of re-emerging disease, especially in Indonesia. Diphtheria can be prevented by immunization. Diphtheria immunization has drastically reduced mortality and susceptibility to diphtheria, but it is still a significant childhood health problem. This study predicted the number of diphtheria patients in several regions using support vector regression (SVR) combined with the genetic algorithm (GA) for parameter optimization. The area is grouped into 3 clusters based on the number of cases. The proposed method is proven to overcome overfitting and avoid local optima. Model robustness tests were carried out in several other regions in each cluster. Based on the experiments in three scenarios and 12 areas, the hybrid model shows good forecasting results with an average mean squared error (MSE) of 0.036 and a symmetric mean absolute percentage error (SMAPE) of 41.2% with a standard deviation of 0.075 and 0.442, respectively. Based on experiments in various scenarios, the SVR-GA model shows better performance than others. Compares two- means tests on MSE and SMAPE were given to prove that SVR-GA models have better performance. The results of this forecasting can be used as a basis for policy-making to minimize the spread of diphtheria cases
Leveraging machine learning for column generation in the dial-a-ride problem with driver preferences
Get PDFThe dial-a-ride problem (DARP) is a significant challenge in door-to-door transportation, requiring the development of feasible schedules for transportation requests while respecting various constraints. This paper addresses a variant of DARP with time windows and drivers’ preferences (DARPDP). We introduce a solution methodology integrating machine learning (ML) into a column generation (CG) algorithm framework. The problem is reformulated into a master problem and a pricing subproblem. Initially, a clustering-based approach generates the initial columns, followed by a customized ML-based heuristic to solve each pricing subproblem. Experimental results demonstrate the efficiency of our approach: it reduces the number of the new generated columns by up to 25%, accelerating the convergence of the CG algorithm. Furthermore, it achieves a solution cost gap of only 1.08% compared to the best-known solution for large instances, while significantly reducing computation time
Machine learning for global trade analysis: a hybrid clustering approach using DBSCAN, elbow, and SOM
Get PDFGlobal trade constitutes a highly complex and interdependent system influenced by diverse economic, geographic, and political factors. This study proposes a hybrid clustering framework that integrates density-based spatial clustering of applications with noise (DBSCAN), elbow, and self-organizing maps (SOM) methods to uncover latent structures in international trade patterns. Utilizing averaged trade data from 25 countries spanning the period from 2013 to 2023, the framework identifies distinct clusters based on export-import characteristics. The DBSCAN is employed to detect dense trade hubs and outlier behaviors, the elbow method determines the optimal number of clusters, and SOM facilitates the visualization of non-linear, high-dimensional trade relationships. The analysis reveals three prominent trade clusters: Global Trade Leaders, Emerging Trade Powers, and Niche Exporters, each reflecting varying degrees of trade diversification and dependency. These empirical findings align with established economic theories, including the Heckscher Ohlin model and dependency theory, and provide actionable insights for policymakers seeking to enhance trade competitiveness and regional integration strategies
Educational data mining approach for predicting student performance and behavior using deep learning techniques
Get PDFEducational Data Mining (EDM) uncovers insights from large datasets collected from various educational platforms, such as online learning systems, student information databases, and classroom tools. EDM helps educators identify hidden patterns that improve teaching strategies, personalize learning experiences, and predict student performance. Predicting student success has become a key focus of EDM, allowing institutions to implement targeted interventions and personalized support. The dataset included academic achievement grades from 1,001 students enrolled in various courses during the fall semester across multiple years, to demonstrate how proposed models provide more accurate predictions compared to traditional machine learning methods. Models such as YOLO, Fast R-CNN, Artificial Neural Networks (ANNs), and Long Short-Term Memory (LSTM) networks are used to capture complex, non-linear relationships within the data. The comparative analysis shows that these deep learning models significantly outperform traditional techniques, such as decision trees and support vector machines (SVMs). The results indicate that proposed method offers improved predictive accuracy, enabling educational institutions to identify at-risk students and deliver tailored interventions. This study highlights the potential of enhanced method to transform personalized education and enhance student success by better understanding individual learning needs and behaviors