International Journal of Artificial Intelligence and Education Technology
2835-2432ISSN (Online)
This study examines the theoretical, methodological, and applied aspects of artificial intelligence (AI) integration into accounting systems and tax administration using expert evaluation methods. A systematic review of contemporary academic literature is conducted to clarify the conceptual foundations of artificial intelligence and to substantiate its functional role in the digital transformation of accounting and fiscal processes. The analysis focuses on key applied areas of AI utilization, including the automation of accounting and managerial operations, the implementation of intelligent accounting information systems, and the use of virtual assistant platforms for tax compliance and automated reporting. Selected applied solutions—such as Robotic Process Automation (RPA), accounting software robots (RobBee), and the virtual assistant DavrOn—are examined to demonstrate the practical potential of intelligent technologies in financial and tax management. The results indicate that AI-based solutions contribute to higher operational efficiency, improved data reliability, enhanced procedural transparency, and greater adaptability of managerial decision-making under conditions of increasing data complexity.
Read MoreDoi: https://doi.org/10.54216/IJAIET.040101
Vol. 4 Issue. 1 PP. 01-07, (2025)
Student retention in higher education institutions is a critical problem that causes academic and financial challenges to individual students and to schools and entire countries. The field of study should be in the area of student retention as it enables educational facilities to provide appropriate intervention. The present study implements a comparative analysis of five machine learning classifiers; Linear Discriminant Analysis, K-Nearest Neighbours, Support Vector Machine, Random Forest and Gradient Boosting classifiers on dataof 4424 students who were selected from the Realinho et al. (2022) data set which contains demographic and socioeconomic, and macroeconomic and academic performance data from a Portuguese higher education institution over a decade. The mutual information feature selection step reduces the 22-dimensional feature space prior to model trainingby selecting 12 features that have, statistically, the highest discriminative power. Five-fold stratified cross-validation shows that the best overall performance is achieved by a SVM with a radial basis function kernel with accuracy of 97.1% and F1 score of 0.954 and all five models achieve AUC greater than 0.981. The importance analysis reveals that the combination of four measures of academic success from the first two semesters constructs 87.6% of the signal that Random Forest model uses for prediction which is driven by the most important predictor - number of curricular units that the student passes during the secondsemester (importance= 0.335). The impact of all socioeconomic and demographic and macroeconomic factors is less than 13%. The findings of the study have three implications about risk factors in student retention via empirical measurement.
Read MoreDoi: https://doi.org/10.54216/IJAIET.040102
Vol. 4 Issue. 1 PP. 08-22, (2025)
Automated Essay Scoring (AES) technologies have been extensively researched for holistic, topic-specific scoring, but their use to predict multiple analytic writing quality traits of English Language Learner (ELL) student essays has received less attention. This research contributes to this knowledge gap by systematically investigating multi-trait AES on the ELLIPSE corpus (Learning Agency Lab, 2022), a publicly accessible dataset of 6,482 argumentative essays written by grades 8-12 ELLs and rated by human raters on six analytic traits: cohesion, syntax, vocabulary, phraseology, grammar, and conventions. We experiment with five models: Ridge regression, Support Vector Regression (SVR) with a radial basis function (RBF) kernel, Random For-est, fine-tuned BERT-base-uncased and fine-tuned DeBERTa-v3-base. The mean Quadratic Weighted Kappa (QWK) across six traits is highest for DeBERTa-v3 (0.726) - a 26.5-point improvementover the Ridge base-line (0.461) and a 6-point improvement over BERT (0.666). Phraseology is the most difficult trait to score automatically (DeBERTa QWK = 0.701) and cohesion the easiest (DeBERTa QWK = 0.742). Analysis of inter-trait correlations reveals high co-variation between vocabulary and phraseology (r = 0.79), which may reflect common linguistic skills that can be leveraged by multi-task learning. Thisresearch sets a replicable baseline for multi-trait AES on the ELLIPSE corpus, and suggests that phraseology scoring is the most urgent area for future architectural innovation.
Read MoreDoi: https://doi.org/10.54216/IJAIET.040103
Vol. 4 Issue. 1 PP. 23-35, (2025)
Modern higher education campuses now use intelligent learning tools as standard educational resources yet students learning results depend on their understanding of these tools and their implementation in academic work. The study analyzes how students pre-pare to use educational tools while investigating the connection between their preparedness and their judgment of educational benefits. The study uses an open student-perception dataset to conduct empirical research which includes developing constructs and profiling readiness and creating predictive models and establishing pathways. The study introduces two measurement methods which include source breadth to measure how students acquire knowledge about intelligent tools through different information channels and an advantage score to present perceived benefits for educational activities. The three-profile segmentation method shows that different groups in the sample display distinct levels of preparedness and value assessment. The Random Forest model demonstrates superior performance because it achieves the highest accuracy among all tested models in the predictive stage. The selected model exhibits an accuracy rate of 0.789 and a precision rate of 0.714 and a recall rate of 1.000 and an F1 score of 0.833 and an area under the receiver operating characteristic curve of 0.806 in hold-out evaluation. The analysis of variable importance indicates that AI knowledge and grade-point average and information breadth and profile membership serve as the main factors that explain the results. The final stage of the process transforms analytical results into distinct educational pathways which focus on developing essential literacy skills and implementing structured curriculum materials and providing support for governance matters and enabling advanced collaborative learning. The results demonstrate that the educational benefits of intelligent tools depend more on students’ preparedness to use them than on their initial exposure to the tools.
Read MoreDoi: https: // doi. org/ 10. 54216/ IJAIET. 040104
Vol. 4 Issue. 1 PP. 36–49, (2025)
The detection of students who will face academic difficulties or leave their studies during their initial course period provides universities with a brief time frame to develop effective solutions. This research paper conducts a systematic analysis which tests multiple machine learning classifiers on the Open University Learning Analytics Dataset (OULAD) which serves as one of the most widely used public educational datasets that presents data from 32593 students who studied 22 different courses through distance learning. The four classification methods include logistic regression decision tree random forest and gradient boosting which use a feature set that combines student demographic information and virtual learning environment (VLE) clickstream-based engagement data. The primary discovery shows that VLE behavioral characteristics constitute the most important elements for Random Forest which identifies total click volume and active VLE days and typical daily click volume as its top four elements which make up 92.8% of total importance while demographic information has less impact. Random Forest achieves the strongest held-out test performance (AUC = 0.998, F1 = 0.978, accuracy = 98.2%) while Decision Tree shows lower results with AUC = 0.959 which demonstrates how performance losses occur when systems need to be understandable. At-risk students in the two groups present a 75.8% decrease in total VLEclicks which results in an average of 49.0 clicks compared to 203.0 clicks with a t value of 104.0 and a p value less than 0.001. The research describes its complete end-to-end prediction pipeline which includes details about its model evaluation framework and its dataset to enable future researchers to reproduce the study. The results have direct implications for the design of early-alert systems and the ethical deployment of predictive analytics in higher education.
Read MoreDoi: https://doi.org/10.54216/IJAIET.040105
Vol. 4 Issue. 1 PP. 50–66, (2025)