Machine Learning Applications in Emergency Management
dc.contributor.advisor | Dong, Zhijie | |
dc.contributor.author | Liu, Dixizi | |
dc.contributor.committeeMember | Valles Molina, Damian | |
dc.contributor.committeeMember | Fulton, Lawrence | |
dc.date.accessioned | 2021-07-22T17:37:19Z | |
dc.date.available | 2021-07-22T17:37:19Z | |
dc.date.issued | 2021-08 | |
dc.description.abstract | Emergency prediction and management are characterized by high dynamics and complexity, and inaccurate prediction and inefficient management can result in the loss of human lives and substantial environmental and economic consequences. Traditional methods for emergency management, such as linear regression and time series analysis, have limitations in handling large-scale data and conducting in-depth analysis. Machine learning (ML) is a branch of artificial intelligence, which plays a vital role in emergency management through modeling and predicting with high accuracy and efficiency. A novel coronavirus disease 2019 (COVID-19) has killed and infected millions of people around the world since late 2019. Controlling the spread of COVID-19 pandemic is a very important and emergent topic in the United States. Moreover, the number of mass shootings in the United States has risen sharply in 2020 under the COVID-19 pandemic. Therefore, in this thesis, we explore ML models to improve emergency management by focusing on two different types of emergency, coronavirus pandemic (i.e., COVID-19) and mass shootings. For COVID-19, we focus on exploring the evolution algorithm and ML to model the effect of social distancing on the spread of COVID-19. Deep Neural Networks (DNN) form a powerful deep machine learning model that can process unprecedented volumes of data. The hyperparameters of DNN have a major influence on its prediction performance. Evolutionary algorithms (EAs) form a heuristic-based approach that provides an opportunity to optimize deep learning models to obtain good performance. Therefore, we propose an evolutionary deep learning model called IPSO-DNN based on DNN for prediction and improve Particle Swarm Optimization (IPSO) algorithm to optimize the kernel hyperparameters of DNN in a self-adaptive evolutionary way. In the IPSO algorithm, not only a micro population size setting is introduced to improve the search efficiency of the algorithm, but also the generalized opposition-based learning strategy is used to guide the population evolution. In addition, the IPSO employs a self-adaptive update strategy to prevent the premature convergence and then improves the exploitation and exploration parameter optimization performance of DNN. In Part Ⅰ, we show that the IPSO provides an efficient approach for tuning the hyperparameters of DNN with saving valuable computational resources. We explore the proposed IPSO-DNN model to predict the effect of social distancing on the spread of COVID-19 based on mobility and social distancing metrics. The preliminary experimental results reveal that the proposed IPSODNN model has the least computation cost and yields better prediction accuracy results when compared to the other comparison models. The experiments of the IPSO-DNN model also illustrate that aggressive and extensive social distancing interventions is crucial to help slow the spread of the COVID-19 epidemic in the United States. For mass shooting, we concentrate on predicting the future number of mass shooting incidents in the United States based on the public’s attitudes on Twitter. In recent years, social media plays a prominent and very important role in the spread of mass shooting incidents and brought about a significant contagious effect on future similar incidents. Therefore, we propose a self-excited contagion model based on sentiment analysis of Twitter data on mass shootings. We explore different ML models to forecast the change in the public’s attitudes over time. These ML models include Support Vector Machine (SVM), Logistic Regression (LR), and the proposed IPSO-DNN model. The performances of different ML models are critically examined based on performance measures such as precision, recall, and accuracy. The results present that the proposed IPSO-DNN model has the significant capability to forecast the changes in public attitudes towards mass shootings on Twitter over time. The proposed self-excited contagion model is to predict the future number of mass shootings by focusing on the magnitude of influence of mass shootings and the spread of public attitudes on Twitter. experiments indicate that the positive attitude plays an important role in analyzing and predicting future similar mass shooting incidents. Especially, due to the economic recession and people's huge pressures related to the lockdowns, the COVID-19 pandemic has significantly increased the number of mass shootings in 2020. Therefore, we also improve the proposed self-excited contagion model with the consideration of social distancing and the daily growth rate of COVID-19 cases to predict and analyze mass shootings under the COVID-19 pandemic. Experimental results of Part Ⅱ demonstrate that our proposed contagion models perform very well in predicting the future mass shootings in the United States. | |
dc.description.department | Engineering | |
dc.format | Text | |
dc.format.extent | 106 pages | |
dc.format.medium | 1 file (.pdf) | |
dc.identifier.citation | Liu, D. (2021). Machine learning applications in emergency management (Unpublished thesis). Texas State University, San Marcos, Texas. | |
dc.identifier.uri | https://hdl.handle.net/10877/14057 | |
dc.language.iso | en | |
dc.subject | machine learning | |
dc.subject | emergency management | |
dc.subject | deep neural networks | |
dc.subject | particle swarm optimization | |
dc.subject | COVID-19 social distancing | |
dc.subject | mass shootings | |
dc.subject | ||
dc.subject | contagion model | |
dc.subject | sentiment analysis | |
dc.title | Machine Learning Applications in Emergency Management | |
dc.type | Thesis | |
thesis.degree.department | Engineering | |
thesis.degree.discipline | Engineering | |
thesis.degree.grantor | Texas State University | |
thesis.degree.level | Masters | |
thesis.degree.name | Master of Science |
Files
Original bundle
1 - 1 of 1