IAES International Journal of Artificial Intelligence (IJ-AI)
Not a member yet
1769 research outputs found
Sort by
Accuracy of long short-term memory model in predicting YoY inflation of cities in Indonesia
Get PDFOur research evaluates the effectiveness of the long short-term memory (LSTM) model in forecasting annual year-on-year (YoY) inflation across 82 cities in Indonesia based on time series data from BPS economic reports for 2014-2024. This study tests the accuracy of the model in reconstructing past inflation patterns, then evaluates the capabilities and limitations of the model in various urban area contexts with the root mean square error (RMSE), mean absolute percentage error (MAPE), and coefficient of determination(R2) metrics. The findings show that LSTM performs well in metropolitan areas such as Jakarta, Bandung, and Surabaya with R2values >0.8 and the lowest MAPE of 10.91% in Jakarta. However, in small cities with higher economic volatility such as Tanjung Pandan, the model shows significant prediction errors (R²<0.50 and MAPE up to 283.11%). Moderate performance (0.50≤ R²≤0.80) was found in cities such as Palembang, Semarang, and Makassar, reflecting the model's adaptive ability to moderate inflation patterns. These results emphasize the important role of structured economic data in improving the reliability of predictions, so that the policy implications of this study include the use of the LSTM model as an early warning system by fiscal and monetary authorities, as well as the need for a data-based inflation control strategy to strengthen regional and national economic resilience in supporting sustainable development towards Indonesia Emas 2045
Federated deep learning intrusion detection system on software defined-network based internet of things
Get PDFThe internet of things (IoT) and software-defined networks (SDN) play a significant role in enhancing efficiency and productivity. However, they encounter possible risks. Artificial intelligence (AI) has recently been employed in intrusion detection systems (IDSs), serving as an important instrument for improving security. Nevertheless, the necessity to store data on a centralized server poses a potential threat. Federated learning (FL) addresses this problem by training models locally. In this work, a network intrusion detection system (NIDS) is implemented on multi-controller SDN-based IoT networks. The interplanetary file system (IPFS) FL has been employed to share and train deep learning (DL) models. Several clients participated in the training process using custom generated dataset IoT-SDN by training the model locally and sharing the parameters in an encrypted format, improving the overall effectiveness, safety, and security of the network. The model has successfully identified several types of attacks, including distributed denial of service (DDoS), denial of service (DoS), botnet, brute force, exploitation, malware, probe, web-based, spoofing, recon, and achieving an accuracy of 99.89% and a loss of 0.005
Artificial intelligence-powered smart roads: leveraging orange3 for traffic signs recognition
Get PDFTraffic sign recognition systems are an important concern of advance driver assistance systems (ADAS) and intelligent autonomous vehicles. Recently, many studies have emerged that aim to employ artificial intelligence (AI) and machine learning (ML) to detect and classify traffic signs to improve a system that can be embedded in vehicles to increase efficiency and safety. This work's primary goal is to address traffic sign identification and recognition utilizing a 2,339-image open-source dataset from Kaggle. Our detection model for extracting and classifying traffic sign suggestions is built using Orange3 data mining tools, based on four classifiers random forest (RF), k-nearest neighbors (KNN), decision tree (DT), and adaptive boosting (AdaBoost). Signs are classified into eight categories: don't go signs, go signs, horn signs, roundabout signs, danger signs, crossing signs, speed limit sign, and unallowed signs. The results of examining and evaluating the proposed model based on the performance evaluation metrics showed that RF outperformed with an accuracy rate of 99.8%, followed by AdaBoost with a classification accuracy of 99.2%, and the classification accuracy of DT and KNN was 98.3% and 94.9%, respectively
Survey on 3D biometric traits for human identification
Get PDFIndividuals are verified and identified using Biometric technology based on their biological or behavioral traits. Biometric-based personal authentication systems are more reliable and user friendly, overruns the traditional personal authentication systems. The physiological biometric traits get abraded due to aging and massive work, while the behavioral biometric traits are having high variations due to external factors such as fatigue, and mood. Among the physiological biometric traits, Finger geometry patterns are widely deployed authentication system reason being its stability, user acceptability and uniqueness. Recent trends in Biometrics attempt to incorporate 3D domain traits, 3D reconstruction is done using 2D multiple images. 3D images are usually more robust and illumination invariant as compared to their 2D counterparts. 3D reconstruction algorithms are compared by finding mean square error (MSE)
A novel fuzzy logic based sliding mode control scheme for non-linear systems
Get PDFSliding mode control (SMC) has been widely used in the control of non-linear systems due to many inherent properties like superposition, multiple isolated equilibrium points, finite escape time, limit cycle, bifurcation. This research proposes super-twisting controller architecture with a varying sliding surface; the sliding surface being adjusted by a simple single input-single output (SISO) fuzzy logic inference system. The proposed super-twisting controller utilizes a varying sliding surface with an online slope update using a SISO fuzzy logic inference system. This rotates sliding surface in the direction of enhancing the dynamic performance of the system without compromising steady state performance and stability. The performance of the proposed controller is compared to that of the basic super-twisting sliding mode (STSM) controller with a fixed sliding surface through simulations for a benchmark non-linear system control system model with parametric uncertainties and disturbances. The simulation results have confirmed that the proposed approach has the improved dynamic performance in terms of faster response than the typical STSM controller with a fixed sliding surface. This improved dynamic performance is achieved without affecting robustness, system stability and level of accuracy in tracking. The proposed control approach is straightforward to implement since the sliding surface slope is regulated by a SISO fuzzy logic inference system. The MATLAB/Simulink is used to display the efficiency of proposed system over conventional system
A web-based learning platform to assess student performance using online session activity engagement
Get PDFPredicting students' performance and engagement is crucial for academic eLearning partners in colleges and universities as well as students themselves considering post-COVID-19 pandemic and university grant commission (UGC) dual degree regulation era. An educational system's data on students’ engagement in taking courses that are a significant component of an institution of higher learning with a cogent vertical syllabus can be used to make predictions. By examining how closely a student's course-taking actions correspond with the requirements of the syllabus, one can utilize the student's conduct in the classroom and online eLearning web tool as a predictor of future achievement. This paper presents a study that uses an eLearning web-based dataset to predict students' success throughout a series of online interactive sessions. The dataset records how students engage with each other during online lab work, including how many keystrokes they make, how long they spend on each task, and how well they perform on exams overall. The current methods lacks accuracy to assess student performance and engagement with high precision. In addressing this paper introduces novel multi-label ensemble learning (MLEL) using XGBoost (XGB) and K-fold cross validation. Experiment outcome shows the proposed (MLEL-XGB) achieves much improved outcome than other existing models
Artificial intelligence for individuals with disabilities in higher education institutions: a systematic review
Get PDFWith the growing integration of artificial intelligence (AI) in education, its potential to support students with disabilities in higher education remains significant but underexplored. This systematic review synthesizes existing literature on AI's effectiveness, barriers, and implications for inclusive education. Using the sample, phenomenon of interest, design, evaluation, and research type (SPIDER) framework, studies published between 2013 and 2024 were identified through a systematic search in databases such as PubMed, Scopus, Embase, Cochrane Library, and Google Scholar. Eighteen studies met the inclusion criteria, focusing on higher education settings and students with disabilities. The findings emphasize AI's role in enhancing accessibility, personalizing learning experiences, and fostering inclusiveness. However, persistent challenges include technological barriers, ethical concerns, and insufficient training. While AI holds transformative potential to support students with disabilities in higher education, addressing infrastructure gaps and ethical and training deficiencies is crucial for sustainable implementation and equitable learning environments
Parametric optimization of microchannel heat exchanger using socio-inspired algorithms
Get PDFMiniaturized products and systems have emerged as game-changing innovations with huge potential in the modern period with increasing emphasis on sustainable development and green energy. Automotive, astronomical, electronics, and medical research are just a few of the industries where micro electro mechanical systems (MEMS) have found use. In addition to that, microchannel heat exchangers (MCHX) have been created in response to the growing demand for effective cooling solutions for these small systems. Optimization of these MCHX is important for improving the overall system efficiency. In this work, two popular socio inspired evolutionary algorithms viz. teaching learning-based optimization (TLBO) and cohort intelligence (CI) are applied for optimizing three objectives such as power density, compactness factor, and heat transfer with pressure drop (HTPD) for air-water MCHX. The results obtained are significantly improved when compared with genetic algorithm (GA). Moreover, both the techniques are observed to be robust. This study investigates the use of socio-inspired artificial intelligence (AI) algorithms to support the design and optimization of heat exchangers, highlighting their potential to address complex engineering challenges more efficiently
Support vector machine performance: simulation and rice phenology application
Get PDFIn the case of classification, model accuracy is expected to result in correct predictions. This study aims to analyze the performance of two kinds of support vector machine (SVM) methods: the support vector machine one versus one (SVM OvO) method and the generalized multiclass support vector machine (GenSVM) method. This method will compare to the generalized linear model, namely the multinomial logistic regression (MLR) method. Simulations were conducted using SVM OvO and GenSVM methods to get an overview of the parameters affecting both methods' performance. Furthermore, the three classification methods are implemented in the case of modelling the rice phenology and tested for performance. Simulation results show that, however, the SVM OvO and GenSVM machine learning methods are sensitive to the choice of model parameters. The empirical study results show that the SVM OvO and GenSVM methods can produce satisfactory model accuracy and are comparable to the MLR method. The best rice phenology model accuracy was obtained from the SVM OvO model, where 79.20 ± 0.21 overall accuracy and 70.69 ± 0.29 kappa were obtained. This research can be continued by handling samples, especially when class members are a minority, and can also add random effects to the SVM model
Large language models for pattern recognition in text data
Get PDFLarge language models (LLMs) are widely deployed in settings where both reliability and efficiency matter. We present a calibrated, seed‑robust empirical comparison of an encoder fine‑tuned model (bidirectional encoder representations from transformers (BERT)‑base) and a decoder in‑context model (generative pre-trained transformer (GPT)‑2 small) across Stanford question answering dataset v2.0 (SQuAD v2.0) and general language understanding evaluation (GLUE)-multi-genre natural language inference (MNLI), Stanford sentiment treebank 2 (SST‑2). Beyond accuracy, we assess reliability (expected calibration error with reliability diagrams and confidence–coverage analysis) and efficiency (latency, memory, throughput) under matched conditions and three fixed seeds. BERT‑base yields higher accuracy and lower calibration error, while GPT‑2 narrows gaps under few‑shot prompting but remains more sensitive to prompt design and context length. Efficiency benchmarks show that decoder‑only prompting incurs near‑linear latency/memory growth with k‑shot exemplars, whereas fine‑tuned encoders maintain stable per‑example cost. These findings offer practical guidance on when to prefer fine‑tuning versus prompting and demonstrate that reliability must be evaluated alongside accuracy for risk‑aware deployment