International Journal of Artificial Intelligence & Mathematical Sciences

Inconsistency Detector between Low-Quality Video and Audio Using Deepfake

2025-08-09T07:42:08+00:00

This study presents the challenges and issues within low-resolution video and audio through the application of sophisticated deep learning methodologies, addressing the prevalent issue of manipulated media in recent times. By employing dual-stream convolutional architecture, we analyze the intricate relationship between auditory and visual cues, commencing with an extensive examination of existing detection approaches and their constraints when confronted with substandard video content. Utilizing the VidTimit dataset as our base, we train and test our model and check its performance in comparison to existing pre-trained models. Our evaluation framework includes accuracy metrics, confusion matrices, and F1 score to ensure efficiency. Using a variety of filters such as Edge Preserving and Gaussian Blur on video data preprocessing, we enhance the detection of disparities by optimizing the input data. Model integration is the hallmark of our innovative dual-stream convolutional architecture, where audio and visual components are perfectly integrated. In this architecture, the visual stream applies convolutional layers to capture spatial characteristics from low-quality video frames, and the audio stream applies RNNs to capture temporal patterns in audio signals. The fusion module effectively integrates these streams and makes way for synchronization analysis and anomaly detection. In this aspect, our trained network does very well in active video detection and lip-reading tasks.

Analyzing Pakistan Textile and Clothing Export trade dynamics- A comprehensive Study from 2003 to 2021 Using Statistical Model

2025-08-09T07:42:08+00:00

Pakistan's textile industry is the core of the nation's economy, employing over 15 million people and contributing over 60% of all export earnings. The research work done in this paper includes the collection of yearly data (2003-2021) of textile exports of Pakistan with other countries. By using MINITAB software Moving Average Model and Residual Plots has been obtained 100% throughout model and it gives appropriate forecast values. This statistical model provides the expected value for the prediction of future textile export data. The main aim of the manuscript is to throw light on dynamic performance of exports of textiles of Pakistan in the perspective of various countries like Afghanistan, Bangladesh, Turkey, Germany, Australia, Sri Lanka, China, India and Qatar.

Student Academic Performance Prediction Using Machine Learning Approach

2025-08-09T07:42:08+00:00

The academic performance of students in higher education has been a focal point of extensive research aimed at addressing issues such as academic underachievement, increased dropout rates, and delays in graduation, among other persistent challenges. Basically, student performance is the accomplishment of educational objectives, be they short-term or long-term. The study therefore describes the prediction of Student Performance aimed at improving the attainment of academic outcomes in higher education. By considering several performance factors like patterns of attendance, parent involvement, individual study habits, and preference in teaching methods, this system can portray a view of each student’s academic performance in a holistic way. Support Vector Machine (SVM) and Artificial Neural Network (ANN) algorithms have implemented in this proposed model to classify and predict student academic performance. A comparative study was also undertaken to determine which of these two algorithms has better accuracy or precision. The results show that the SVM outperformed the ANN.

Artificial Intelligence and Statistical Regression for the Prediction of Temperature over Sukkur Region

2025-08-19T07:57:13+00:00

This study focuses on forecasting the temperature of the Sukkur region in Sindh, Pakistan, using historical temperature data from four neighboring cities: Kashmore, Shikarpur, Ghotki, and Khairpur. Three different predictive models were developed, based on multiple regression, supervised machine learning, and artificial neural networks (ANN). The results indicate that all three approaches provide highly accurate temperature predictions, with multiple regression and supervised machine learning performing slightly better than the ANN model. The analysis is based on temperature data from 2001 to 2019, and all simulations were conducted using Python 3.9.16 within the Anaconda environment.

E-Lock: A Blockchain Framework for Enhancing Security and Trust in E-Learning

2025-08-19T07:56:40+00:00

This paper introduces E-Lock, a blockchain-based framework aimed at enhancing security, scalability, and trust in e- learning
systems. By leveraging Polygon’s blockchain platform, E- Lock utilizes decentralized ledgers and smart contracts to improve transparency,
efficiency, and data security in digital education environment. Polygon’s high-throughput, cost-effective architecture overcomes limitations in
traditional e-learning platforms, enabling scalable, low-cost transaction management. Through decentralized consensus, cryptographic hashing,
and interoperability with the Ethereum ecosystem, E-Lock ensures data integrity, verifiability, and secure operations while reducing
vulnerabilities to malicious attacks. The framework empowers educational institutions to maintain data sovereignty, protect intellectual property
rights, and foster resilience and transparency in the e-learning ecosystem. This paper provides practical insights and guidance for researchers in
the field of Technology-Enhanced Learning (TEL), presenting a comprehensive analysis of the benefits and trade-offs associated with
blockchain integration in educational platforms.

Next Generation Blockchain Framework for Educational Document Security

2025-08-19T07:56:14+00:00

The handling and verification of educational documents pose significant challenges for students, employers, and
institutions. Students face difficulties managing physical documents for admissions, job applications, and attestations, while
organizations struggle to verify their authenticity. Current processes are slow, error-prone, and vulnerable to fraud. This paper
introduces a blockchain-based digital identification and verification system that leverages a decentralized ledger to ensure
data integrity, security, and accessibility. Unlike traditional and cloud-based systems, our solution eliminates third-party
dependencies, offering an immutable, tamper-proof platform. Educational institutions can input student data linked to unique
national identification numbers, enabling seamless certificate and experience verification while prioritizing user privacy.
Shorter verification times and increased stakeholder trust are the outcomes of a prototype implementation. However, adoption
costs and scalability remain limitations. This solution lowers fraud, speeds up processes, promotes environmental
sustainability, and establishes the foundation for secure, scalable credential management globally through paperless
operations.