Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Muhammad Shujah Islam Sameem has a total of 9 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Evaluating Low-Light Image Enhancement for Object Detection

    by Abdulrahman Abdulaziz Almudaymigh, M. Shujah Islam Sameem

  • Open Access

    Personality-Aware Course Recommender System Using Deep Learning for Technical and Vocational Education and Training

    by Rana Hammad Hassan, Malik Tahir Hassan, Muhammad Shujah Islam Sameem, Muhammad Aasim Rafique

    Personality represents enduring patterns, providing insights into an individual’s aptitude and behavior. Integrating these insights with learning tendencies shows promise in enhancing learning outcomes, optimizing returns on investment, and reducing dropout rates. This interdisciplinary study integrates techniques in advanced artificial intelligence (AI) with human psychology by analyzing data from the trades of Technical and Vocational Education and Training (TVET) education, by combining them with individual personality traits. This research aims to address dropout rates by providing personalized trade recommendations for TVET, with the goal of optimizing outcome-based personalized learning. The study leverages advanced AI techniques and data from a nationwide TVET program, including information on trades, trainees’ records, and the Big Five personality traits, to develop a Personality-Aware TVET Course Recommendation System (TVET-CRS). The proposed framework demonstrates an accuracy rate of 91%, and a Cohen’s Kappa score of 0.84, with an NMAE at 0.04 and an NDCG at 0.96. TVET-CRS can be effectively integrated into various aspects of the TVET cycle, including dropout prediction, career guidance, on-the-job training assessments, exam evaluations, and personalized course recommendations.


  • Open Access

    Link prediction grounded upon complex networks based model, based on MJMI-RW

    by Khush Bakhat, Kashif Kifayat, M Shujah Islam, M Mattah Islam

    Link prediction is the process of predicting the connection probability between two nodes based on observable network data, such as network structural topology and node properties. Despite the fact that traditional similarity-based methods are basic and effective, their generalisation performance varies greatly across networks. In this paper, we propose a novel link prediction approach, MJMI-RW, based on the Maxwell Jüttner distribution endowed by the Mutual Information, which recovers the probability of a node's connection by applying node characteristics to a system with less computation. Initially, the method investigates a comprehensive node feature representation by combining diverse structural topology information with node importance properties through feature construction and selection. The selected node features are then fed into a supervised learning task that solves the features matrix using the node features as input. The enhancements of MJMI-RW in terms of the average area under the curve and the precision of state-of-the-art algorithms compared to the finest baseline networks when compared to baseline methods. The limitation of MJMI-RW is its minimal computational complexity feature construction and substitution of complex features with semantic node attributes. Moreover, since inductive matrix completion is a supervised learning task in which the underlying low-rank matrix can be solved by representative nodes instead of all their nodes, it offers a potential link prediction solution for large-scale networks.

  • Open Access

    Employing FGP-3D, a Fully Gated and Anchored Methodology, to Identify Skeleton-Based Action Recognition

    by M Shujah Islam, Abdullah Algosaibi, Warda Rafaqat, Khush Bakhat

    Recent years have seen an explosion in interest in and development of action recognition based on skeletal data. Contemporary methods using fully gated units can successfully extract characteristics from human skeletons by relying on the human topology that has been predefined. Despite advancements, fully gated unit-based techniques have trouble generalizing to other domains, particularly when dealing with various human topological structures. In this context, we introduce FGP-3D, a novel skeleton-based action recognition technique that can generalize across datasets while being effective at learning spatiotemporal features from human skeleton sequences. This is accomplished via a multi-head attention technique to learn an ideal dependence feature matrix from the uniform distribution. We next re-evaluate state-of-the-art techniques as well as the suggested novel descriptor FGP-3D in order to examine the cross-domain generalizability of skeleton-based action recognition in real-world video skeleton statistics. After being applied to commonly used action categorization datasets, experimental results demonstrate that the proposed FGP-3D, with pre-training, generalizes well and outperforms the state-of-the-art.