Joel de Goma

Joel de Goma

Auto-generated profile, based on publicly available information

Joel de Goma

Mapua University, School of Information Technology, Makati, Metro Manila, Philippines

Research Focus & Metrics

Academic Career Length

5Years (2022-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
12 January 2024
Verified
Conference Paper
Open access
JGJoel de Goma
Joel de Goma
MRMarcus Julius Rabano
Marcus Julius Rabano
DSDean Marlou Sanson
Dean Marlou Sanson
KTKerstein Alyssa Tadena
Kerstein Alyssa Tadena
This study focuses on the detection of counterfeit currency within the modern Philippine banknotes, specifically, the Php 500 and Php 1000 denominations of the New Generation Currency series. To address this issue, we propose an innovative solution that leverages the combination of MobileNetv2 and YOLOv5s, employing transfer learning for efficient feature detection. Our research involves a comprehensive evaluation of MobileNetv2's performance, comparing it to transfer learned ResNet-18 as a feature extractor, and the optimization of hyperparameters of the former model. We assess seven critical security features, while aiming to enhance counterfeit currency detection efficiency. Our methodology encompasses dataset acquisition, data augmentation, image annotation, and model customization. The results of our study demonstrate that MobileNetv2 achieves a mean Average Precision (mAP) of 92.8% for detecting 500-denomination banknotes and 93% for 1000-denomination banknotes, all while maintaining lightweight computational requirements. In contrast, ResNet-18 achieves a higher mAP of 93.6% and 95%, respectively, but at the cost of heavyweight computational demands. This research offers a promising and practical approach to significantly improve counterfeit currency detection in the Philippine banking industry by employing security feature detection technique.

11 September 2024
Verified
Conference Paper
Open access
JGJoel C. De Goma
Joel C. De Goma
JOJamie Beatrice L. Olivar
Jamie Beatrice L. Olivar
HPHobie L. Perez
Hobie L. Perez
SVSadie Catherine E. Velasco
Sadie Catherine E. Velasco
AVAlberto C. Villaluz
Alberto C. Villaluz
In this study, the researchers evaluate the performance of five deep learning models—ResNet-34, ResNet-50, VGG-19, ViT, and a Custom Model combining ResNet-50 and EfficientNet-B0—to discern the authenticity of digital artworks. The Grad-CAM heatmap is applied to the mentioned models except for ViT, as its architecture relies on self-attention mechanisms rather than convolutional layers, posing challenges for its application. Specifically, the researchers’ goal is to distinguish between artworks created by human artists and those generated by Artificial Intelligence (AI), thereby contributing to the fair recognition of human creativity in art. The researchers' analysis offers insights into the methodology employed and the comparative performance of these models based on various metrics. Notably, the custom hybrid model outperforms the others, achieving an impressive accuracy of 92%. In contrast, ResNet-34, ResNet-50, VGG-19, and ViT models demonstrate accuracies of 85%, 51%, 76%, and 53%, respectively. This comparison highlights the Custom Model's effectiveness in discerning the origin of artworks, indicating its potential as a valuable tool for distinguishing between AI-generated and authentic pieces, thereby assisting in ensuring the genuineness of human artistic creativity.

23 February 2024
Verified
Conference Paper
Open access
JGJoel De Goma
Joel De Goma
JBJohn Adam Bravo
John Adam Bravo
SPSpringtime Prudente
Springtime Prudente
RRRobert Francis Rondilla
Robert Francis Rondilla
The spread of SMS spam messages has grown as a major issue, offering a chronic annoyance to mobile phone users worldwide. These unwanted SMS, sent with malevolent purpose or to promote a company, have evolved into increasingly complex schemes. In response to this expanding threat, this article sets out to improve existing SMS spam detection models using deep learning methodology and word embedding techniques. The researchers will emphasize the importance of combating SMS spam and the history of spam filtering, as well as detail the study technique, which includes data pretreatment such as punctuation removal, lowercase conversion, word stemming, and TF-IDF vectorization. The experimental framework employs both LSTM and BiLSTM models, both with and without TF-IDF vectors, for a total of four unique models. Each model is subjected to rigorous K-Fold cross-validation, with the results comparing the efficacy of TF-IDF vectorization in boosting SMS spam detection. This study aims to provide mobile phone users with improved defenses against the harmful threat of SMS spam.

13 September 2024
Verified
Conference Paper
Open access
JGJoel de Goma
Joel de Goma
VIVicente Julius Delfin Iii
Vicente Julius Delfin Iii
EFEzekiel Joshua Flores
Ezekiel Joshua Flores
JTJose Edward Tabuton
Jose Edward Tabuton
AVAlberto Villaluz
Alberto Villaluz
— The purpose of this research is to start a study that would eventually create a framework that would be a guide in the administration and interpretation of HFDT. In this paper, the Researchers proposed a model using Cascade object detectors and the Viola Jones algorithm. This model would identify and describe the facial features of human sketch from the human figure drawing test (HFDT) based on the image dataset that the researchers gathered through random sampling. The accuracy for facial feature identification and description were 31% and 35% respectively after the first test. After adjustments were made, Identification accuracy went up to 86% and Description accuracy went up to 82%. This paper will also discuss on how this model can be improved further.

14 October 2022
Verified
Conference Paper
Open access
LBLenz Baron S. Balita
Lenz Baron S. Balita
KDKyle Matthew A. Degrano
Kyle Matthew A. Degrano
APAndrei Daniel A. Pamoso
Andrei Daniel A. Pamoso
JGJoel C. De Goma
Joel C. De Goma
Sentiment analysis is one of the most prominent methods on gathering and analyzing insightful textual data from various sources. The information produced from such a method can be imperative in understanding the general public's sentiment on a certain product or service. Over the years, countless sentiment analysis models have already been established using known algorithms such as Naive Bayes, Support Vector Machine, and many more. However, with the advent of novel technologies and neural networking, recent studies have employed Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM) Recurrent Neural Network together to formulate more efficient and modernized models (Rehman, Malik, Raza, & Ali, 2019). As such, the study proposed a similar model to analyze the sentiments of book user reviews from GoodReads categorized according to three distinct genres – children's, young adults’, and romance. The paper also aimed to determine the viability and effects of amalgamating features such as Word2Vec, POS, and SenticNet to the overall accuracy (Ayutthaya & Pasupa, 2018). Once the model was trained to the procured dataset, the results suggested that combining Word Embedding, POS, and SenticNet features drastically improves its performance in contrast to other tested variations. Amalgamating the three features to a CNN-LSTM hybrid model yielded an F1-score of 90%; whilst other variants with lacking features or a standalone CNN or LSTM model only resulted to F1-scores around 86% below. Graphing the performance of all the constructed models to an ROC curve also indicated the effectiveness of the proposed model – having an AUC value of 0.9588.

26 April 2024
Verified
Conference Paper
Open access
JGJoel de Goma
Joel de Goma
RMRosselle Mae Mercado
Rosselle Mae Mercado
TSTimothy Jacob Saporna
Timothy Jacob Saporna
SVSean Paolo Vallo
Sean Paolo Vallo
— This study presents an in-depth analysis of sentiment expressed in tweets related to vaccines using Natural Language Processing techniques particularly Bidirectional Encoder Representations from Transformers or BERT. With the increasing relevance of vaccines in public health discussions and the growing impact of social media on public opinion, understanding the sentiment surrounding vaccines is of paramount importance. The researchers collected a diverse dataset of tweets discussing vaccines and employed BERT, a powerful deep learning model that perform sentiment analysis. The model got an outstanding result with an accuracy of 89.77%, meaning that it correctly predicted the class labels for 89.77% of the instances in dataset. Additionally, the AUC value of the model is 0.9438 which is close to a value of 1. This shows that the model can distinguish between two classes in the dataset. Sentiment analysis of tweets towards vaccines using BERT can provide valuable insights into public perception and help health authorities and policymakers make informed decisions. It's also essential to approach this task with care, considering the nuances of sentiment in social media data.