ABSTRACT
One such protocol currently enforced by the Philippine government to combat COVID-19 is the mandatory use of face masks in public places. The problem however is that ensuring people follow this protocol is difficult to monitor during a pandemic due to other conflicting health protocols like social distancing and workforce reduction. This study therefore explores on the creation of deep learning models that consider both frontal and side view images of the face for face mask detection. In doing so, improvements to robustness were found when compared to using models that were previously trained on purely frontal images. This was accomplished by first relabeling a subset of images from the FMLD dataset. These images were then split into train, validation, and test sets. Four deep learning models (YOLOv5 Small, YOLOv5 Medium, CenterNet Resnet50 V1 FPN 512x512, CenterNet HourGlass104 512x512) were later trained on the training set of images. These four models were compared with three models (MobileNetV1, ResNet50, VGG16) that were trained previously on purely frontal images. Results show that the four models trained on the relabeled FMLD dataset offer an approximately 20% increase in classification accuracy over the three models that were previously trained on purely frontal images. Copyright © 2022 by SCITEPRESS - Science and Technology Publications, Lda. All rights reserved.
ABSTRACT
Since the start of the pandemic in early 2020, there have been numerous studies related to the design and use of disease models to aid in understanding the transmission dynamics of COVID-19. Output of these models provide pertinent input to policies regarding restricting or relaxing movements of a population. Perhaps the most widely used class of models for COVID-19 disease transmission is the compartmental model. It is a population model that assumes homogeneous mixing, which means that each individual has the same likelihood of contact with the rest of the population. Inspite of this limitation, the approach has been effective in forecasting the number of cases based on simulated scenarios. With the shift from nationwide lockdowns to granular lockdown as well as gradual opening of limited face to face classes, there is a need to consider other models that assume heterogeneity as reflected in individual behaviors and spatial containment strategies in smaller spaces such as buildings. In this study, we use the COVID-19 Modeling Kit (COMOKIT, 2020) as a basis for the inclusion of individual and spatial components in the analysis. Specifically, we derive a version of COMOKIT specific to university setting. The model is an agent-based, spatially explicit model with the inclusion of individual epidemiological and behavior parameters to show evidence of which behavioral and non-pharmaceutical interventions lead to reduced transmission over a given period of time. The simulation environment is set up to accommodate the a) minimum number of persons required in a closed environment including classrooms, offices, study spaces, laboratories, cafeteria, prayer room and bookstore, b) parameters on viral load per building or office, and c) percentage of undetected positive cases going on campus. The model incorporates the following interventions: a) compliance to health protocol, in particular compliance to wearing masks, b) vaccine coverage, that is, the percentage distribution of single dose, two doses and booster, c) distribution of individuals into batches for alternating schedules. For mask compliance, as expected, results showed that 100% compliance resulted to lowest number of cases after 120 days, followed by 75% compliance and highest number of cases for 50% compliance. For vaccine coverage, results showed that booster shots play a significant role in lowering the number of cases. Specifically, those who are fully vaccinated (2 doses) and 100% boosted produce the lowest number of cases, followed by the 50% of the population fully vaccinated and have had their booster shots. Intervals of no onsite work or class in between weeks that have onsite classes produce the lowest number of cases. The best scenario is combining the three interventions with 100% compliance to mask wearing, 100% fully vaccinated with booster, and having two batches or groups with interval of no onsite classes. © 2022, The Author(s), under exclusive license to Springer Nature Switzerland AG.
ABSTRACT
As most governments in the world currently face the pandemic, various policies and initiatives have been put in place in order to help control the spread of the COVID-19 outbreak. While these initiatives and interventions are taking place, a pandemic still creates a reality of risk and uncertainty. In these kinds of situations, public trust is greatly important to properly mitigate health and societal impacts of the pandemic. Social media platforms could be utilized as sources of information to gain insight on public sentiment, especially with the rise of social media utilization during the quarantine [13]. Given this, the study attempts to analyze social media sentiments particularly found in Twitter in order to not only look into the polarity of public sentiment on the government, its processes, and its policies, but particularly, to detect trust between the governed and the ones governing. Furthermore, it seeks to examine and analyze the trust narratives present in the Philippines currently. In this study, a supervised machine learning model was created using Linear SVC, utilizing TF-IDF and n-grams for feature extraction and selection in order to detect the respective trust category of a given sentiment and predict the trust category of new data points. While the results are overall negative, examining the trust categories individually demonstrates different narratives that dictate, affect, and express citizen trust towards different aspects of the government. The behavioral trust group provided narratives on certain political figures involved in a string of anomalies for the negative category, while the positive category lauded the VP for her continued service amidst the pandemic. On the other hand, narratives in the institutional trust group revolved around national and local institutions, where talks about national institutions being more prominent in the negative category, while local institutions, such as local government units, are found in the positive category. Lastly, narratives on the operational trust group focused on certain pandemic policies (lockdowns, mass testing, contact tracing) for the negative side, while vaccines and vaccinations were the focus for the positive side. © 2022, The Author(s), under exclusive license to Springer Nature Switzerland AG.
ABSTRACT
As of January 02, 2022, the Philippines is combating another surge in COVID-19 cases. With vaccinations still ongoing, the country remains vigilant and the government continues to promote compliance to minimum health standards as preventive measures to minimize the spread. Disinformation remains a challenge especially if compliance to minimum health standards and adoption of health interventions are necessary to curb the spread of COVID-19. Incorrect and unverified information about the virus increased as well which continues to run rampant in social media and with minimal models to detect disinformation in a Philippine context. The study aimed to understand the features of disinformation of COVID-19 in a Philippine context with the goal of creating a text classification model to detect disinformation of COVID-19 in social media to promote vaccine usage in the country. The usage of social network analysis was performed to understand the narratives present regarding COVID-19 disinformation. Words related to vaccines, government corruption, and government mismanagement were prevalent under the disinformation categories of “False” and “Mostly False” while words related to health information such as cases or vaccine counts were prevalent under the “Mostly True” and “True” category. Linear SVM text classification model performed the best through accuracy, precision, and recall in detecting disinformation by using TF-IDF as a feature compared to using both TF-IDF and n-grams. Disinformation narratives revolved around the idea of COVID-19 cases/vaccines, government mismanagement, and regulations. Results showed that disinformation caused distrust of the government’s management over the pandemic. Moreover, the spread of disinformation was contained to the user itself and spread to at least one other user. © 2022, The Author(s), under exclusive license to Springer Nature Switzerland AG.
ABSTRACT
Social media can be used to understand how the public is responding to the ongoing nationwide COVID-19 vaccination campaign, allowing policymakers to respond effectively through informed decisions. However, conducting social media analysis in the Philippine-context presents a challenge because natural informal conversations make use of a combination of English and local language. This study addresses this challenge by including part-of-speech tags, frequency of code switching and language dominance features to represent bilingualism in training machine learning models with COVID-19 vaccination-related Tweets for sentiment and emotion analysis. Results showed that the English-Tagalog Logistic Regression sentiment classification model performed better than Textblob, VADER and Polyglot with an accuracy of 70.36%. Similarly, the English-Tagalog SVM emotion classification model performed better than Text2emotion, NRC Affect Intensity Lexicon and EmoTFIDF with an average mean-squared error of 0.049. The added bilingual features only improved these performance metrics by a small margin. Nevertheless, SHAP analysis still revealed that sentiment and emotion classes exhibit varying levels of these bilingual features, which shows the potential in exploring similar linguistic features to distinguish between classes better during text classification for future studies. Finally, Tweets from September 2021 to January 2022 shows negative, mainly anger and sadness, perceptions towards COVID-19 vaccinations. © 2022, The Author(s), under exclusive license to Springer Nature Switzerland AG.
ABSTRACT
After the World Health Organization declared COVID-19 a global pandemic in 2020, vaccination was seen as a major intervention in transmission reduction and in achieving herd immunity. However, initial response of the public to vaccination has been met with uncertainties. Moreover, while the first world countries have increased mobility due to vaccination, the Philippines have yet to cover more regions in the next few months. With YouTube being the most popular video-based social media platform for seeking information, this study explored emotions expressed by the general public among COVID-19 vaccine-promoting, vaccine-neutral and vaccine-discouraging YouTube videos. NRC Word-Emotion Association Lexicon was used to identify the emotions expressed in the video comments from the three video-tone categories. The W-ANOVA and Games-Howell post-hoc results showed that vaccine-promoting videos have significantly higher anticipation, joy, and surprise emotion scores, while sadness and fear emotion scores are significantly higher in vaccine-discouraging videos. Furthermore, trust emotion score is significantly high in vaccine-neutral videos. Understanding the basic emotions expressed by viewers on vaccine-related videos may serve as a guide in crafting effective health promotion campaigns and could provide insights on other complex emotions related to vaccination. including hesitancy and envy. © 2021 IEEE.
ABSTRACT
Coronavirus disease (COVID-19) is an infectious disease, which is caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) that was identified in December 2019 in Wuhan, China [1], [2]. It is a pandemic that causes respiratory disorder and is transmitted through sneezing droplets of infected individuals. These droplets can fall on the objects around the effected and enter a healthy individual through contact. Major symptoms of this disease include lethargy, dry cough, followed by fever [3]. The number of cases is surging dramatically, raping developed and undeveloped countries together [3]. According to the World Health Organization (WHO) COVID-19 weekly epidemiological Update for 29th of December there are 79 million infected cases and 1.7 million deaths globally. This pandemic not only affects our health but also affects our livelihood. In the absence of specific treatment or a vaccine, non-pharmaceutical interventions (NPI) form the backbone of the response to the COVID-19 pandemic. These NPI includes physical distancing, regular hand washing, and wearing a face mask. This study aims to help with the monitoring of these NPIs specifically wearing face masks using deep learning. This study implements face mask detection and recognition system that automatically detects and recognizes if a person is wearing a Medically approved face mask, Non-Medically approved face mask, or not wearing a mask at all. This study has determined that MobileNetV1 model has shown the best performance regarding classification (79%) and processing speed up to 3.25 fps. © 2021 IEEE.
ABSTRACT
Mathematical models of the COVID-19 pandemic have been utilized in a variety of settings as a core component of national public health responses. Often based on systems of ordinary differential equations, compartmental models are commonly used to understand and forecast outbreak trajectories. In view of the primarily applied nature of COVID-19 models, theoretical analysis can provide a global and long-term perspective of key model properties, and relevant insights about the infection dynamics they represent. This work formulates and undertakes such an investigation for a compartmental model of COVID-19, which includes the effect of interventions. More specifically, this paper analyzes the characteristics of the solutions of a compartmental model by establishing the existence and stability of the equilibrium points based on the value of the basic reproductive number R0. Our results provide insights on the possible policies that can be implemented to address the health crisis. © 2021 American Institute of Physics Inc.. All rights reserved.
ABSTRACT
Contact-tracing is part and parcel of interventions in reducing the rate of transmission of the disease. After identifying the rate of transmission, the next important step is to determine the spread of the disease using network analysis. Using manual or digital methods, active tracing requires testing of suspect cases to identify and isolate positive cases from identified close contacts. Passive tracing allows citizens to report symptoms to designated local health authorities. Most contact tracing efforts implemented by the local government units to stem the transmission of COVID-19 follow standard manual contract tracing procedures. However, with the rapid increase in the speed of the outbreak, traditional contact tracing approaches are not sufficient to contain the spread of the disease. FASSSTrace is designed to make contact tracing more efficient by developing a contact tracing model for digital platforms. Specifically, the platform provides a social network model of confirmed (C) and suspect (S) cases and visualizes the transmission dynamics with the inclusion of suspect cases. The method allows for the construction of a model reflecting the contact network of the confirmed cases as recorded in the official disease surveillance tool which produces a contact network determining superspreaders from key individuals and locations. he contacts network uses a two-mode network incorporating geographical locations as nodes to bridge the unlinked confirmed cases. Analysis involves micro-level network measures to determine key individuals and locations and macro-level network measures to study the patterns in transmission dynamics of the disease. © 2021 ACM.