Donata Acula

Dr. Donata Acula

University of Santo Tomas
orcid

Research Focus & Metrics

Academic Career Length

10Years (2017-Present)

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

Representative Publications

View All Publications
1 May 2017
Verified
Conference Paper
Closed access
RCReymond Joseph A. Cabrera
Reymond Joseph A. Cabrera
CLCriselle Amor P. Legaspi
Criselle Amor P. Legaspi
EPErika Jasmine G. Papa
Erika Jasmine G. Papa
RSReden D. Samonte
Reden D. Samonte
Donata AculaDonata Acula
Donata Acula
Different approaches in Image Processing are widely used in the field of medicine and since health is the factor being considered, it must be vigilantly done. This study aims to develop a system that will apply the different techniques of Image Processing and Genetic Algorithm, specifically in automating the detection of Leukemia. The benchmarks in detecting Leukemia that are being used by the medical experts on a blood sample structure are taken into account. The researchers will create a system that will accept a blood image and process it using the said methods and added a Naïve Bayes Classifier in the detection process. A total of 100 blood images are used in evaluating the system. The system achieved a detection rate of 90.33%.

1 April 2018
Verified
Conference Paper
Closed access
GAGillian Lindsay V. Alquisola
Gillian Lindsay V. Alquisola
DCDaniel Jose A. Coronel
Daniel Jose A. Coronel
BRBryan Matthew F. Reolope
Bryan Matthew F. Reolope
JRJulia Nicole A. Roque
Julia Nicole A. Roque
Donata AculaDonata Acula
Donata Acula
This study is all about predicting and visualizing the disaster risks in the Philippines brought by tropical cyclones. According to UNICEF Philippines (2017), the Philippines is highly exposed to natural hazards because it lies along the Pacific Typhoon Belt compounded by uncontrolled settlement in hazard-prone areas, high poverty rate, and failure to implement building codes and construction standards. The number of casualties resulting from these incidents can be reduced if the possible occurrence of risk can be foreseen which can help the community have awareness and form recommendations ahead. The researchers of this study improved the results of Autoregressive Integrated Moving Average (ARIMA) and Artificial Neural Network (ANN) by using a de-noising model of Discrete Wavelet Transform (DWT), in predicting the disaster risk levels of the Philippine provinces in terms of casualties, damaged houses, and damaged properties. Based on the result of the study, the best model for casualty, damaged houses, and damaged properties is the DWT-ARIMA-ANN model.

30 March 2021 in Journal of Modeling and Simulation of Materials
Verified
Journal Article
Open access
PDF file
Donata AculaDonata Acula
Donata Acula
This paper employed the intelligent approach based on machine learning categorized as base and ensemble methods in classifying the disaster risk in the Philippines. It focused on the Decision Trees, Support Vector Machine, Adaptive Boosting Algorithm with Decision Trees, and Support Vector Machine as base estimators. The research used the Exponential Regression for missing value imputation and converted the number of casualties, damaged houses, and properties into five (5) risk levels using Quantile Method. The 10-fold cross-validation was used to validate the proposed algorithms. The experiment shows that Decision Trees and Adaptive Decision Trees are the most suitable models for the disaster data with the score of more than 90%, more than 75%, more than 75% in all the classification metrics (accuracy, precision, recall f1-score) when applied to classification risk levels of casualties, damaged houses and damaged properties respectively.

1 January 2018
Verified
Conference Paper
Closed access
YMYosuke R. Matsuoka
Yosuke R. Matsuoka
GSGabriel Angelo R. Sandoval
Gabriel Angelo R. Sandoval
LSLuis Paolo Q. Say
Luis Paolo Q. Say
JTJann Skyler Y. Teng
Jann Skyler Y. Teng
Donata AculaDonata Acula
Donata Acula
In this technological age, handwriting communication is still an essential aspect in the lives of people and relating to each other. This study was created to identify the most suitable set of algorithms that can be used and determine how effective it would be in recognizing cursive handwritten texts. The proponents created a system that accepts a handwritten text image as input, undergoes processing stages and outputs a text based on the features extracted per character using the Diagonal Feature Extraction, and classification using Euler Number with the use of the Modified One-Pixel Width Character Segmentation Algorithm. A total of 100 handwritten text images are used in evaluating the system. The system achieved a character recognition rate of 88.7838% and word recognition rate of 50.4348%.

1 April 2018
Verified
Conference Paper
Closed access
Donata AculaDonata Acula
Donata Acula
LOLouise Aster C. Oblan
Louise Aster C. Oblan
TPTracy B. Pedroso
Tracy B. Pedroso
KRKatrine Jee V. Riosa
Katrine Jee V. Riosa
MTMichelle Arianne R. Tolibas
Michelle Arianne R. Tolibas
In the technology age, articles with fraudulent content are rampant, especially articles shared on social media. Misinformation could just be an inaccuracy at its best, or it could lead to normalizing false information at worst. To aid the predicament, the researchers created a system that will “fact check” suspicious articles against those articles that have been deemed credible, reliable, and more accurate, in order to help fight deceiving content that may be detrimental to society. The journal regarding computational fact checking that was published by Ciampaglia, et. al. (2015) from the Indiana University in the USA entitled Computational Fact Checking from Knowledge Networks, was used as the basis and inspiration for this thesis. The researchers made use of the undirected graph (UG) together with a part-of-speech (POS) tagging algorithm to create a knowledge graph (KG) that would serve as the center of the system. Five different POS tagging algorithms were paired with the UG to assess which combination would yield the best results, these are Conditional Random Fields, Logistic Regression, a Hybrid of CRF and LR, Random Forests, and K-Nearest Neighbors. Random Forests and K-Nearest Neighbors were classification algorithms used in Ciampaglia's study. It was concluded that among the 5 pairs of UG and POS Tagging algorithms, the Hybrid of CRF and LR used as a POS tagger, together with the UG, created the most efficient KG.

20 February 2025 in RADIOELECTRONIC AND COMPUTER SYSTEMS
Verified
Journal Article
Closed access
PDF file
JJJunidar Junidar
Junidar Junidar
MMMelinda MELINDA
Melinda MELINDA
DDDinda D. Diannuari
Dinda D. Diannuari
Donata AculaDonata Acula
Donata Acula
ZZZulfan Zainal
Zulfan Zainal
The subject of this paper is the detection of Autism Spectrum Disorder (ASD) traits using facial recognition based on thermal images. The goal of this study was to evaluate and compare the performance of various Convolutional Neural Network (CNN) architectures in classifying thermal facial images of children with ASD, thereby facilitating the early identification of autistic traits. The tasks addressed include preprocessing a dataset of thermal facial images to prepare them for model training; conducting classification using three CNN architectures VGG-19, ResNet50V2, and EfficientNet; and assessing their performance based on accuracy, precision, recall, and F1-score metrics. The methods employed involved training these CNN models on a balanced dataset of 4,120 thermal facial images and splitting them into training, validation, and test sets. Each model underwent extensive training to determine its ability to effectively classify autism and non-autism classes. The results revealed that ResNet50V2 achieved the highest accuracy of 98.82%, followed by VGG-19 and EfficientNet with accuracies of 96.47% and 96.07%, respectively. ResNet50V2 also demonstrated superior generalizability due to its lower validation loss and higher classification accuracy compared to other architectures. Conclusion. The scientific novelty lies in: 1) introducing thermal imaging as an effective tool for detecting ASD traits; 2) demonstrating the superior performance of ResNet50V2 in classifying thermal facial images with high accuracy and generalization; and 3) exploring EfficientNet for the first time in this domain, highlighting its potential for improving autism diagnostic systems. This study contributes to advancing noninvasive methods for ASD detection and paves the way for further applications of deep learning in clinical diagnostics.