Support Vector Machine Hyperparameters Experiment using Human Eye Datasets

 Ernest Glenn M. Villanueva, Kristian Paulo P. Flores, Louise Denzel L. Granadoz, and Rolando B. Barrameda

College of Science and Computer Studies, De La Salle University – Dasmariñas, Dasmariñas, Cavite

ABSTRACT

Drowsy driving contributes for about 9.5% of all car accidents. Researchers have focused on the eyes as a potential biometric for detecting sleepiness, with the hope of reducing the frequency with which such accidents occur. Although SVM and other classification algorithms have been used to distinguish between closed and open eyes in a number of relevant studies, very little work has been done to optimize their parameters via testing. We built the SVM model for this research project using Python, and we used and analyzed datasets involving the human eye. In order to determine which hyperparameter will provide the best accuracy, experiments were carried out. We found that the SVM model with the RBF hyperparameter had the best accuracy (99.53%) when compared to other models that used other hyperparameters (regularization, gamma, and kernel).

Keywords: vector, support vector machine, SVM, accuracy, drowsiness, image processing, image classification

Aletheia: A Web-Based Tool for Detecting Fake News using Algorithmic Benchmarking

Vince L. Dela Peña, Elnathan John L. Salavarria, and Kynch Tito R. Silao

College of Engineering, Architecture and Technology, De La Salle University – Dasmariñas, Dasmariñas, Cavite

ABSTRACT

The continuous evolution of the world wide web paved way for faster dissemination of information throughout the world through different domains such as news outlet websites and social media. Though this makes information accessible within the tips of fingers, this also includes the danger of misinformation or dissemination of false information. To address this, creating an automated fact checker is found as a solution. This study focused on creating this solution using machine learning algorithms to provide a fact checking tool using supervised learning. The models created from the algorithm were used in a web application to evaluate inputs such as Text and URL. All the components were handled on the cloud using Amazon Web Services as the primary cloud hosting provider. An Application Program Interface was used to access the model and predict. The study led to the development of the Aletheia Combination that uses Ensemble Learning Classifier to combine three algorithms namely Decision Tree, Neural Network, and Random Forest. The Aletheia Combination provided a 95.38% F1-Score in the Train-Test-Split Validation and an accuracy of 76.36% when it comes to real world testing.

Keywords: Machine Learning, Algorithms, API, Ensamble Learning, Decision Tree, Neural Network, Random Forest, Train-Test-Split Validation

Endepedia: Wildlife Image Recognition and Educational Mobile App using YOLOv4 Convolutional Neural Networks

Lincoln Morales, Jeff Joecel Punzalan, Philip Nicole Vergara, Jospehine Eduardo, and Rolando Barrameda

College of Science and Computer Studies, De La Salle University – Dasmariñas, Dasmariñas, Cavite

ABSTRACT

This study aims to develop an application to detect YOLO4 and classify animals using Convolutional Neural Networks. The application provides quiz features to test the user knowledge and provide information about nearby location where the animals can be found. The results show that the trained model achieved an accuracy of 93.91% using mAP@IoU of 0.50 and 86.75% using the YOLOv4-tiny. Moreover, 90% of the respondents concur that the application is reliable in terms of successfully launching the application, and 88% of the respondents were able to use the application to its intended objective. Also, the occurrence of crashes were determined to be minimal to none. Endepedia is a robust mobile application created to classify animals using machine learning and is able to carry out its purpose of educating users about endemic and endangered animals by providing access to a wide range of information and means of learning.

Keywords: Machine Learning, Algorithms, YOLOv4, educational mobile app, convolutional neural network

Utilizing the Research Findings for Enhanced Implementation of Outcomes-Based Education Practices and Curriculum Design in Blended Learning Settings

Marivic Mitschek^1,2 and Rosanna Esquivel²

¹College of Science and Computer Studies, De La Salle University – Dasmariñas, Dasmariñas, Cavite

²Angeles University Foundation Graduate School

ABSTRACT

The modern workplace is swiftly digitizing, demanding students to be prepared for future careers. In response, education has shifted towards blended learning and outcomes-based education, aiming to equip students for the dynamic digital age. This transformation poses challenges, requiring adaptability, innovation, and a lifelong learning mindset from educators. The COVID-19 pandemic has further accelerated online learning, highlighting the necessity to align assessments with intended learning outcomes. This study employed data mining to evaluate this alignment and the impact of learning tasks on student outcomes focusing on course type and delivery both course and topic levels. The findings hold potential to benefit educators, students, administrators, and contribute to workforce development and curriculum improvement. Both course learning outcomes and assessment level outcomes across classes effectively targeted both lower-order and higher-order thinking skills. However, a notable misalignment was observed in topic learning outcomes, which predominantly focused on lower-order thinking skills. This underscores the necessity to address this misalignment in the course design framework.

Keywords: blended learning, outcomes-based education, assessment alignment, data mining, student outcomes

JOURNAL OF ENGINEERING, COMPUTING AND TECHNOLOGY

The Journal of Engineering, Computing and Technology (JECT) is a scholarly peer-reviewed, and academic research journal for engineers, computer scientist, academicians, and research scholars aiming to publish innovative, state of the art results coming from research and advances in different aspects of engineering, architecture, computer science and information technology.  JECT is dedicated to the dissemination of advanced technologies that will be beneficial to professionals and academic researchers. It is a bi-annual publication that releases issues every June and December.

The journal is governed by an editorial board whose members specialize in each of the mentioned disciplines.  The editor in chief is elected by the members of the editorial board.

Editorial Board

Editor-in-Chief

Maryli F. Rosas, DIT (Computer Science, De La Salle University-Dasmariñas)

Associate Editor

Ma. Cristina A. Macawile, PhD (Engineering, De La Salle University- Dasmariñas)

Managing Editor

Mr. Jaime Zeus C. Agustin (University Research Office, De La Salle University-Dasmariñas)

Members

Paulino H. Gatpandan, DIT (Computer Science, De La Salle University- Dasmariñas)

Maryjoie A. Lituanas, MSES (Engineering, De La Salle University- Dasmariñas)

Mengvi P. Gatpandan, DIT (Computer Science, Jose Rizal University, Mandaluyong City)

Daniel Dasig Jr., PhD (Field, Change Management Engineer, TELUS Communication Inc., Canada)

Nestor Tiglao, PhD (Electrical and Electronics Engineering Institute, University of the Philippines – Diliman)

Atty. Rudolph Val F. Guarin (Field, FGNG Law Office)

Prof. Sonia M. Pascua (Field, Drexel University, United States of America)

Brojo Kishore Mishra, PhD (Department of CSE, School of Engineering, GIET University, Gunupur, India)

Valentina Emilia Balas, PhD (Automation and Applied Informatics, Intelligent Systems Research, University of Arad, Romania)