International Journal of Research in Computing

Technological Innovations and Their Impact on Alumni Engagement and Relationship Management in Higher Education

Jagdeep Sharma — 2026-01-14

The evolving digital ecosystem in higher education institutions (HEIs) has positioned alumni as critical stakeholders in institutional development, mentoring, and academic community building. Despite the growing adoption of digital platforms, empirical evidence on technology-mediated alumni engagement - particularly in the context of Indian HEIs – remains limited, creating a gap in understanding effective digital engagement strategies. This study aims to examine the role of technology-enabled alumni engagement practices in HEIs of Punjab, India, by analysing the relationship between demographic characteristics, digital interaction patterns, and perceived usefulness of technological tools in strengthening alumni–institution relationships. A mixed-method research design was employed, integrating quantitative survey data collected from 60 alumni across multiple HEIs with qualitative insights from two institutional representatives. Quantitative analysis involved descriptive statistics, Chi-square tests, and correlation analysis to evaluate associations among demographic variables, online engagement frequency, and satisfaction with digital tools. Qualitative data were used to contextualize and interpret institutional perspectives. Findings indicate that over 80% of alumni prefer digitally supported engagement mechanisms, with LinkedIn and WhatsApp emerging as the most utilized platforms. Chi-square analysis reveals significant associations between alumni participation and year of graduation, program type, and campus placement history. Correlation results demonstrate a positive relationship between frequency of digital interaction and satisfaction with engagement initiatives. Qualitative evidence highlights strong appreciation for mentorship programs, online seminars, and professional networking, while identifying incomplete alumni databases and limited institutional support as key challenges. The study underscores the strategic importance of digital technologies in enhancing alumni affiliation, professional networking, and intergenerational collaboration. It recommends integrated digital engagement systems, continuous data updating, and personalized communication strategies to foster long-term alumni loyalty and mutually beneficial institutional relationships.

Underwater Waste Detection Using Deep Learning: A Performance Comparison of YOLOv7–10 and Faster R-CNN

Malithi — 2026-01-14

Underwater pollution is one of today’s most significant environmental concerns, with vast volumes of garbage found in seas, rivers, and landscapes around the world. Accurate detection of these waste materials is crucial for successful waste management, environmental monitoring, and mitigation strategies. In this study, we investigated the performance of five cutting-edge object recognition algorithms, namely YOLO (You Only Look Once) models, including YOLOv7, YOLOv8, YOLOv9, YOLOv10, and Faster Region-Convolutional Neural Network (R-CNN), to identify which model was most effective at recognizing materials in underwater situations. The models were thoroughly trained and tested on a large dataset containing fifteen different classes under diverse conditions, such as low visibility and variable depths. From the above-mentioned models, YOLOv8 outperformed the others, with a mean Average Precision (mAP) of 80.9%, indicating a significant performance. This increased performance is attributed to YOLOv8’s architecture, which incorporates advanced features such as improved anchor-free mechanisms and self-supervised learning, allowing for more precise and efficient recognition of items in a variety of settings. These findings highlight the YOLOv8 model’s potential as an effective tool in the global fight against pollution, improving both the detection capabilities and scalability of underwater cleanup operations that will also aid environmental AI specialists and interested parties.

Enhancing E-Commerce Recommendation Systems Using Big Data Analytics by Analyzing Customer Engagement Factors

Gayantha Dilshan — 2026-01-14

ABSTRACT - This research explores the enhancement of e-commerce recommendation systems through Big Data analytics with a particular focus on understanding and leveraging customer engagement factors. The study aimed to identify the key drivers of consumer interaction, assess the impact of integrating Big Data analytics and develop a predictive model to improve the accuracy and relevance of personalized recommendations. A mixed-methods approach was employed, combining quantitative analysis of digital behavioral data with qualitative insights gathered through surveys. This integration enabled the development and robust validation of a Big Data-driven predictive model for personalized e-commerce recommendations. The findings show that combining digital behavioral data with consumer perceptions significantly enhances predictive accuracy as evidenced by improved error metrics (MAE, RMSE), higher explanatory power (R²) and stronger classification outcomes (precision, recall, AUC-ROC). Click frequency, session duration and perceived recommendation accuracy emerged as key predictors of engagement while data triangulation confirmed the model’s reliability. In conclusion, this study demonstrates the value of Big Data-driven personalization in e-commerce, offering practical benefits such as user engagement, marketing efficiency and conversion rates. Academically, it advances knowledge on predictive modeling and recommendation systems, underscoring the effectiveness of mixed methods and advanced analytics. Future research should examine refinements and the long-term impact of dynamic personalization strategies.

INDEX TERMS - Big Data Analytics, Customer Engagement, E-Commerce Recommendation Systems, Personalization, Predictive Modeling

Computational Knightian Uncertainty: Undecidability and the Limits of Cyber Risk Quantification in Software-Intensive Firms

Michel Nguyen — 2026-01-14

Frank Knight’s distinction between measurable risk and unmeasurable uncertainty is central in economics and finance. Contemporary practice often collapses uncertainty into risk by assuming that all material hazards can, in principle, be quantified given sufficient data and computation. This assumption breaks down when the asset in question is large-scale software. This paper argues that undecidability in computability theory, as exemplified by Turing’s halting problem and Rice’s theorem, creates a structural form of uncertainty for software-intensive firms that cannot be reduced to standard probabilistic risk. Many security, safety, and compliance properties of interest to insurers, acquirers, and regulators are non-trivial semantic properties of programs and therefore undecidable in general, even under idealized conditions of perfect code visibility and unlimited classical computation. We call the resulting residual uncertainty computational Knightian uncertainty (CKU): a component of uncertainty that persists even if all observable information is known and arbitrarily robust classical computation is available. We introduce structural opacity the extent to which a codebase resists compression into a small set of regular patterns under a chosen description language [3] and explore approximate Kolmogorov complexity (KC) of codebases as one proxy for this opacity. We develop a conceptual model that links undecidability, structural opacity, and observable outcomes, including cyber incident severity, cyber insurance loss experience, and merger and acquisition (M&A) valuation discounts. In this model, KC and related metrics act as structural covariates that may correlate with CKU, rather than as direct risk measures. Two small synthetic simulations illustrate the empirical logic: first, a crude gzip-based compressibility index sharply separates highly regular from highly irregular synthetic code; second, a KC-like covariate is recoverable in regression when it truly affects incident severity and does not appear systematically when it does not. Our theoretical commitment is modest: computability results guarantee that CKU is non-zero for sufficiently expressive systems. The further claims that CKU is economically material in large, structurally opaque codebases and that structural metrics provide usable proxies are empirical hypotheses to be argued and tested, not consequences of undecidability alone.

An Unsupervised Machine Learning Approach for Dynamic Anomaly Detection and Risk Defense in Cloud Servers

N.W. Chanaka Lasantha — 2026-01-14

This paper presents a hybrid unsupervised machine learning model for real-time anomaly detection and dynamic risk defence in the cloud environment. Traditional security mechanisms such as Intrusion Detection Systems (IDS) and fixed firewall rules are often not sufficient to deal with the emerging threats in cloud computing, especially zero-day exploits and polymorphic malware that evade signature-based detection. The proposed system combines Isolation Forest (IF), Local Outlier Factor (LOF) and Density Based Spatial Clustering of Applications with Noise (DBSCAN) to identify both point and cluster anomalies from unlabelled cloud traffic. Integration with AWS Web Application Firewall (WAF) allows it to update its rules automatically and independently mitigate the threats detected. The model was trained and validated on 2.3 million AWS EC2 traffic records, in addition to the CICIDS2017 dataset which was split into a 70-30 training-validation split. The computation environment consisted of AWS EC2 t2.xlarge (4vCPUs, 16GB RAM) instances of Python 3.8, scikit-learn 0.24.2, MongoDB 4.4, and TensorFlow 2.6. Experiments showed a detection accuracy of 92 per cent with a false positive rate of four per cent. The comparative analysis demonstrated better adaptability and less manual intervention in comparison to traditional IDS (Snort, Suricata: 88 -90% accuracy, 7-9% false positives) and standalone ML models (IF: 87-90% accuracy, LOF: 86 -90% accuracy, DBSCAN: 84 -90% accuracy). The system was able to detect and block port scanning, DDoS, brute-force and data exfiltration patterns in real time with latency of less than 50ms. The reduction of false-positive by 43-56% led to 150-200 alerts per day being reduced in enterprise settings. The hybrid unsupervised model makes the cloud more resilient with adaptive defence without the need for labelled data. Removing manual firewall updates will save 15-20 hours per week for security teams. Future directions are encrypted traffic analysis based on metadata-based behavioural profiling, largescale distributed data processing (10M+ requests/minute), and multi-cloud integration between AWS, Azure, and GCP.