International Journal of Machine Intelligence and Computing

AutoML Algorithms for Predicting Students’ Dropout and Academic Success in Higher Education

2026-06-03T11:38:03+08:00

Challenges with high dropout rates and low academic achievement continue to be a major issue for higher education institutions. To deal with these challenges, it is essential to identify the underlying causes and develop strategic solutions to support student success. Machine learning techniques are powerful tools for predicting student academic performance and enabling timely, targeted interventions. In this study, the goal is to test Automated Machine Learning (AutoML) frameworks that can predict student dropout and academic success. The dataset included students' demographic, socio-economic, and academic information gathered from institutional records. H2O AutoML, TPOT, PyCaret, and AutoGluon were evaluated, as these tools automate important parts of the machine learning pipeline, making it easier to train and select models. Evaluation metrics including accuracy, precision, recall, and F1-score, as well as training time and CPU usage, were used to assess model performance and computational efficiency. Three main experiments were conducted: the first evaluated model diversity, the second observed the effect of cross-validation fold size, and the third assessed different search space configurations. In the first experiment, TPOT obtained the highest accuracy of 90.04%. In the second experiment, using 10-fold cross-validation, H2O AutoML obtained the best accuracy of 91.38%. In the third experiment, TPOT maintained its accuracy at 90.04% across all configurations. A comparison between the conventional ExtraTreesClassifier and TPOT's AutoML ExtraTreesClassifier showed comparable performance, with the conventional model achieving slightly higher accuracy of 90.20% and recall of 94.44%, while TPOT demonstrated superior precision of 90.24%. These results demonstrate that AutoML frameworks are effective tools for early identification of at-risk students, offering competitive performance with significantly reduced manual effort. Future studies could explore deploying these models within academic advising systems.

Quasi-Optimal Low-Complexity Algorithms for the Permutation Flowshop Problem with Weighted Tardiness Penalties

2026-07-15T09:22:52+08:00

The permutation flowshop scheduling problem (PFSP) with N jobs and M machines is NP-hard for M > 2. This paper considers a generalized objective combining a fixed penalty for each tardy job with a penalty proportional to tardiness, under job-specific release times and due dates. Existing heuristics such as NEHedd address the conventional proportional-tardiness objective and do not account for this penalty structure. We propose three low-complexity heuristics: (i) a Pareto-filtered insertion heuristic extending the NEH construction, (ii) a windowed near-exhaustive selection heuristic, and (iii) a hybrid that retains the better solution from the insertion heuristic and NEHedd. The methods are evaluated on 5,100 randomly generated instances spanning three job-to-machine ratios and three traffic intensities. The insertion heuristic achieves 7.2% lower average penalty than selection while running 16 times faster, and its advantage increases from 0.6% at low traffic to 12.5% at high traffic. Insertion achieves lower average penalties than NEHedd on all tested instance sizes under low traffic, and on smaller instances under medium and high traffic. The hybrid achieves over 10% penalty reduction compared to NEHedd for smaller instances, while the relative reduction decreases with increasing configuration size. Runtime analysis indicates polynomial scaling for insertion, with O((N M )b), 1.31 < b < 1.67.

Smart IoT-Based Water Tank Monitoring using ESP2 and Mobile Application

2026-08-12T12:36:55+08:00

In rural, institutional or residential areas, water shortage and ineffective water tank monitoring continue to be major problems leading to unplanned water outages and ineffective water resource management. Conventional monitoring techniques involve manual inspections which are time-consuming, prone to human error and cannot offer real-time information. Therefore, the aim of this research is to create a Smart IoT-Based Water Tank Monitoring System that enables real-time monitoring of water level and TDS-based purity indicators via a mobile application. An ESP32 microcontroller, a waterproof ultrasonic sensor (JSN-SR04T) to measure water level, a Total Dissolved Solids (TDS) sensor to monitor water quality and a DHT11 sensor to measure ambient temperature and humidity are all integrated into the proposed system. Users can remotely access real-time information and receive low water level alarms by using Firebase Real-Time Database as a cloud platform to synchronize sensor data with a mobile application created with Android Studio. A prototype model is used to create a system that facilitates design, implementation, testing and iterative improvements throughout the development process. Meanwhile, functional and real-world tests were conducted to evaluate the sensor accuracy, data transmission reliability, application responsiveness, and system stability under various operating situations. According to the experimental findings, the proposed system effectively tracks water levels and TDS-purity parameters in real time, ensures reliable communication between the hardware and the cloud database, and promptly notifies users when water levels exceed pre-defined criteria. In addition to improving monitoring effectiveness and reducing manual intervention, the created system provides a feasible and affordable alternative for smart water resource management. To accommodate more smart water management applications, further research can improve the system by adding predictive analytics, historical data visualization, and other water quality criteria.

Network Traffic Classification using Machine Learning: A Review of Decision Tree, Random Forest, and SVM Approaches on CIC-IDS2017

2026-03-05T22:17:36+08:00

Cyber-attacks are proliferating rapidly, and networks today are increasingly complex. For this reason, the ability to sort and locate intrusions in network traffic is critical. Researchers have employed methods based on machine learning (ML) as the conventional automated and signature focused detection methods often fail to catch newly emerged attack patterns. The purpose of this work is categorized as the survey is to review state-of-the-art machine learning (ML) network traffic classification techniques, focusing on one specific dataset, i.e., the CIC-IDS 2017 dataset and provides an analysis of today’s research. The study drew on the IEEE Xplore, Springer and Elsevier databases for conducting an extensive review of literature published between 2020 and 2024 having identified studies involving supervised machine learning methods like Decision Trees, Random Forest, Support Vector Machine (SVM) that detected desirable and undesirable traffic. Researchers have also reviewed examples of performance measures such is accuracy, precision, recall and F1-score in different studies to determine areas of strength/weakness or patterns. Studies show that Decision Trees are interpretable, Random Forest has good memory and generalization abilities and SVM is accurate but it requires computational optimization in large datasets. Despite these benefits, challenges such as dataset imbalance, feature redundancy, and real-time implementation continue to remain unsolved. This paper has shown how machine learning can improve the network intrusion detection and challenging issues that need to be addressed in further. The obtained insights lead to a basis for an enhanced scalable and robust traffic classification model, contributing to the goals of current research in this domain.