APPLYING DEEP LEARNING METHODS FOR SHORT TEXT ANALYSIS IN DISEASE CONTROL
Need help with a related project topic or New topic? Send Us Your Topic
DOWNLOAD THE COMPLETE PROJECT MATERIAL
APPLYING DEEP LEARNING METHODS FOR SHORT TEXT ANALYSIS IN DISEASE CONTROL
ABSTRACT
Developing countries have been plagued by recurring outbreaks of infectious diseases; paired with the limitations of traditional disease management tactics, various approaches to disease control have been investigated, with social media at the forefront.
Because the data from this source is brief, noisy, and informal in nature, traditional natural language processing (NLP) approaches are not well suited to its structure. To classify disease-related tweets, deep learning algorithms for character-level word vector learning were investigated, and an adaptive prediction model for epidemic monitoring was constructed using the Ebola virus disease as a case study.
When compared to existing state-of-the-art architectures, our system performed better for the given task; additionally, our predictive model exhibited correlation with officially reported cases, with an early warning of fourteen days prior to official.
Deep learning, NLP, illness control, short text analysis, and word vector learning are all keywords.
CHAPTER ONE: BACKGROUND OF THE STUDY
1.1 General Introduction
The current surge of data has ushered in a new era for human civilisation. Social media platforms such as Facebook, LinkedIn, and Twitter allow people to share information in real time; in 2016,
Africa had 120 million Facebook users per month, and statistics for other social media platforms show similar growth (Fuseware & World Wide Worx, 2014; Parke, n.d.).
The growing use of social media in Africa cannot be overstated, nor can its potential in meaningful projects such as event monitoring, perception evaluation, information extraction and retrieval, and thus its recommendation in disease prevention measures.
Regardless of the success of social media approaches in politics and business, and its prospects in epidemiology as timely, collaborative, and populace-centric; extensive analysis is required,
as information is prone to being misconstrued when machines process natural language. As a result, deep learning methods in social media analytics are required.
1.2 Background Of The Study
Prior to the 15th century, Africa was plagued by epidemic illness outbreaks, which slowed both the region’s human population growth and development aspirations (Spinage & House, 2012).
The study of these cases reveals a trend of illness outbreaks reoccurring in previously affected nations and migrating to surrounding countries, which may be attributable to ecological changes in the region (Kebede, Duales, Yokouide, & Alemu, 2010; Spinage & House, 2012).
12
The recurrent frequency and accompanying mortality rate raises concerns about the amount of readiness, surveillance efficacy, and control efforts in place; hence, numerous ways to mitigating disease propagation have been researched and fused with existing methods over the years.
1.2.i Historical Accounts Of Infectious Disease Outbreaks in Africa
There are several viral infections that are prominent in Africa, including cholera, meningitis, influenza, yellow fever, rickettsia, smallpox, HIV/AIDS, Lassa fever, and Ebola. The total mortality rate is in the millions, with Ebola and HIV/AIDS accounting for more than 3 million deaths (Spinage & House, 2012).
Table 1 provides a sample of disease outbreak cases in Africa, along with the estimated casualty numbers.
Table 1 shows the history of disease epidemics throughout Africa.
1.2.ii Disease control
Disease control, as described by Walter R. Dowdle, is ‘the reduction of disease incidence, prevalence, morbidity, or mortality to a locally acceptable level as a result of purposeful efforts; continuous interventions are required to maintain the reduction’ (Dowdle, 1998).
Disease control measures aim to minimise the contact rate of transmission, maintain a low infectious population, shorten the infection span of the common disease, and achieve disease-free equilibrium (DFE) in the community (Brauer & Castillo-Chavez, 2014).
• Prevention efforts for disease event surveillance, preparedness, and rapid reaction. • Eradication actions for infectious person isolation, treatment, and rehabilitation.
A solid disease control strategy includes both prevention and eradication, however they frequently overlap and can be carried out in different sequences throughout the intervention cycle.
For disease control purposes, a number of organisations are affiliated with the World Health Organisation (WHO), the Centres for Disease Control (CDC), and the health ministries of various nations.
These organisations have been involved in initiatives such as epidemic preparedness and response (EPR) and integrated disease surveillance and response (IDSR).
Adapting traditional data collection, illness identification, epidemic events and projections, casualty estimation, and all other disease outbreak metrics.
1.2.ii.1 Social Media in Disease Control
Traditional methods of information gathering, validation, and dissemination tend to be gradual because to the decision-pipeline involved, which is unavoidable due to the sensitivity of health problems.
Based on the fatality and transmission rates of the recent Ebola virus disease 14 (EVD) and Lassa fever outbreaks in 2014 and 2017, it is only logical to conclude that a quick response would have averted the disaster and decreased the mortality rate.
In recent years, social media in disease control has been investigated to obtain timely information for illness event surveillance, disease prevalence and detection in spatial regions, increasing its potential in disease prediction and prevention (Choi, Cho, Shim, & Woo, 2016).
Unlike documents, which are over 1,000 characters long, formal, and adhere to the syntax of the target language, social media data is comprised of texts that are fewer than 500 characters long and do not adhere to syntax; thus, deep learning approaches are preferred for analysing their structure.
1.2.iii Deep Learning in Short Text Analysis
Word representation by word embedding has been determined to be most appropriate for word similarity and sentence classification tasks in short text analysis (Komninos, 2016), which include sentiment analysis, machine translation, question type classification, topic categorization, and word similarity for web queries and search processing.
Groundbreaking work by Bengio, Ducharme, Vincent, and Janvin (2003), Mikolov, Sutskever, Chen, Corrado, and Dean (2013), and Mikolov, Corrado, Chen, and Dean (2013), which used multilayer convolutional neural networks to capture word semantics and syntactic properties, paved the way for future advances in natural language processing (NLP).
In the event of a disease outbreak, prompt intervention is required to keep death rates under control; this can be accomplished by effective disease monitoring and control.
Only timely and correct curation of information provided via social media and text messaging can predict disease prevalence, and the parameters determined from text data can be combined into statistical models to forecast disease dynamics.
Though the potential with social media data are exciting, the informality of the data creates a high noise level and language processing hurdles.
Do you want to read the entire paper titled Applying Deep Learning Methods For Short Text Analysis In Disease Control? Click here.
Need help with a related project topic or New topic? Send Us Your Topic