NUML International Journal of Engineering and Computing

The Effect of Dimensionality Reduction on Machine Learning Models Performance: A Review

Inam Ul Haq — 2026-01-14

The performance of Machine learning models are fully dependent on the training data. In today’s age where many companies transforming their businesses towards digitalization. A huge volume of raw data is generated and collected, extracting important and useful features from raw data is quite a big challenge, because it is computationally very expensive and time-consuming to train machine learning models on raw data and result in poor model’s performance due to irrelevant features. So, it is necessary to filter out data by reducing the size of features or dimensions and extracting important features from a huge amount of data before implementing any machine learning model on it. In this situation, Dimensionality reduction techniques come into play to eliminate unnecessary or irrelevant features from the data set. In this paper, we are going to review some of the important dimensionality reduction techniques on different machine learning models and evaluate the performance of how dimensionality reduction techniques improve the overall performance of models. After reviewing the research work of different authors, we conclude our results by comparing different implemented techniques of dimensionality reduction, the results before reducing the size of features/dimensions, and after implementing dimensionality reduction techniques and figure out key findings and limitations of reviewed papers.

Conflict of Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Funding

The research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

Data Fabrication/Falsification Statement

The author(s) declare that no data have been fabricated, falsified, or manipulated in this study.

Participant Consent

The authors confirm that Informed consent was obtained from all participants, and confidentiality was duly maintained.

Copyright and Licensing

For all articles published in the NIJEC journal, Copyright (c) of this study is with author(s).

A Review of Real-Time Deep Learning–Based Object Detection Models for Resource-Constrained Embedded Systems

Awais Shah — 2026-01-14

Computer vision relies heavily on object detection; from autonomous drones to industry monitoring it is everywhere. However, despite these advances when one wants to implement these cutting-edge object detection models in embedded system it proves to be difficult due to the limitations in processing power, memory and energy consumption. A detailed examination of real-time object detection models designed for resource-constrained devices is presented in this paper. We investigate widely used one stage detectors like SSD (Single Shot Multi Box) and YOLO (You Only Look Once). In addition, we also discuss model compression methods like knowledge distillation, pruning and quantization, which enable efficient deployment on embedded systems.

Conflict of Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Funding

The research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

Data Fabrication/Falsification Statement

The author(s) declare that no data have been fabricated, falsified, or manipulated in this study.

Participant Consent

The authors confirm that Informed consent was obtained from all participants, and confidentiality was duly maintained.

Copyright and Licensing

For all articles published in the NIJEC journal, Copyright (c) of this study is with author(s).

Vehicle detection and tracking using kalman filter and hungarian algorithm for driving assistance in VANET’s

Rao Umer Farooq — 2026-01-19

In today’s world, the huge increase in automobile vehicles counts on the roads in both rural and urban areas have turned out to create large number of issues which result in the administering and governing of the vehicle on the roads and highways for better driving assistance. Vehicle identification and monitoring by using the information collected from transport monitoring system leads to a defining standard approach for the complex transportation system. In our proposed research Kalman filtering and Hungarian approach used for the identification and tracking of vehicles which mainly focuses on moving vehicles in context of vehicular ad-hoc networks (VANET’s). This paper illustrates the identifying and monitoring the multiple vehicles using deep learning for driving assistance. Kalman filtering is used for monitoring the objects with the pre-trained COCO dataset. Pipeline simplifies the object detection by producing the bounding box around the vehicle. This approach quickly detects the moving vehicle from the running video with the bounding box.

Conflict of Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Funding

The research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

Data Fabrication/Falsification Statement

The author(s) declare that no data have been fabricated, falsified, or manipulated in this study.

Participant Consent

The authors confirm that Informed consent was obtained from all participants, and confidentiality was duly maintained.

Copyright and Licensing

For all articles published in the NIJEC journal, Copyright (c) of this study is with author(s).

A Systematic Literature Review on Modern Cryptographic and Authentication Schemes for Securing the Internet of Things

Tehseen Hussain — 2026-01-25

The rapid integration of the Internet of Things (IoT) into healthcare ecosystems has revolutionized patient monitoring and data accessibility; however, it has simultaneously expanded the cyber-attack surface, leaving sensitive medical data vulnerable to sophisticated breaches. This systematic literature review (SLR) addresses the critical challenge of balancing high-level security with the severe resource constraints of medical sensors and edge devices. By synthesizing evidence from 80 high-impact studies including 18 primary research articles published between 2022 and 2025 this paper evaluates the quality and efficacy of emerging cryptographic frameworks. The methodology utilizes a rigorous quality assessment framework to categorize research into "Strong," "Moderate," and "Weak" tiers. Key findings reveal a significant paradigm shift toward lightweight symmetric ciphers, such as GIFT and PRESENT, and certificateless authentication protocols like ELWSCAS, which reduce communication overhead in narrow-band environments. The analysis further explores the role of blockchain-assisted decentralization and DNA-based encryption in mitigating Single Point of Failure risks and providing high entropy. While decentralized models significantly enhance data integrity, they frequently encounter a scalability wall regarding transaction latency. Furthermore, the review assesses quantum readiness, noting that while lattice-based standards are being ported to microcontrollers, memory footprints remain a barrier for simpler sensors. Ultimately, this SLR maps the current technical frontiers and provides a strategic roadmap for future research, emphasizing the transition toward lightweight, quantum-resistant architectures as the next essential step in securing the global healthcare IoT infrastructure.

Conflict of Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Funding

The research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

Data Fabrication/Falsification Statement

The author(s) declare that no data has been fabricated, falsified, or manipulated in this study.

Participant Consent

The authors confirm that Informed consent was obtained from all participants, and confidentiality was duly maintained.

Copyright and Licensing

For all articles published in the NIJEC journal, Copyright (c) of this study is with author(s).

The Design of Epi-RPNN for the Analysis of Bacteriophage Infection Model

Sabir Ali — 2026-01-29

In this study, we examine the analysis with accuracy based on intelligent computing for the bacteriophage infection model frequently employed in epidemiology. The microbiological is an interesting phenomenon known as bacteriophage infection or phage infection. Bacteriophages are viruses that target and infect bacteria specifically, then use the bacteria as hosts for their own replication. By injecting their genetic material into the bacterial cell, these phages cause the host cell’s machinery to be redirected in order to produce more phages, which ultimately causes the lysis or obliteration of the bacterial host. The derivation of the basic reproduction number and numerical simulations are conducted through a new machine learning approach called the random projection neural network (RPNN) method. The accuracy and robustness of our methodology are examined through a comparison of the results with numerical solvers ode23t and ode15s available in MATLAB. Moreover, the data testing, training and validation of mean square error are examined through performance, training test, error histogram, regression and fitness plots.

Conflict of Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Funding

The research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

Data Fabrication/Falsification Statement

The author(s) declare that no data has been fabricated, falsified, or manipulated in this study.

Participant Consent

The authors confirm that Informed consent was obtained from all participants, and confidentiality was duly maintained.

Copyright and Licensing

For all articles published in the NIJEC journal, Copyright (c) of this study is with author(s).