A Distributed Processing Architecture for Disease Spread Analysis in the

PDSA-RS Platform

Denilson S. Ebling

1 a

, Felipe Machado

1 b

, Glenio Descovi

1 c

, Nicolas Cardenas

2 d

Gustavo Machado

2 e

, Vinicius Maran

1 f

and Alencar Machado

1 g

Laboratory of Ubiquitous, Mobile and Applied Computing (LUMAC), Federal University of Santa Maria,

Roraima Av. 1000, Santa Maria, Brazil

Department of Population Health and Pathobiology, College of Veterinary Medicine, North Carolina State University,

Raleigh, U.S.A.

Keywords:

Intelligent Systems, Decision-Support Systems, Diseases Control, Outreabk Control.

Abstract:

In today’s world, machine learning systems have permeated various domains, from object detection to disease

spread prediction, playing pivotal roles in decision-making processes. Amid the COVID-19 pandemic, the

utilization of machine learning methods like artiﬁcial neural networks and LSTM networks has signiﬁcantly

enhanced forecasting accuracy for disease outbreaks. This paper delves into the development of an intelligent

system proposed by Cardenas et al. (2022a), focusing on simulating disease spread in animals and facilitating

control measures through a stochastic model. Leveraging Docker containers for deployment, this system offers

valuable insights for public health interventions, enabling swift responses to disease outbreaks. The primary

objective of this work is to provide veterinarians with a user-friendly tool that integrates a stochastic model

through an intuitive interface, aiding in critical decision-making processes in a scalable manner. The paper

outlines the background of the stochastic model, introduces the proposed system for integrating and addressing

the identiﬁed problem, presents an evaluation scenario to validate the system’s efﬁcacy, and concludes with

insights drawn from this research endeavor.

1 INTRODUCTION

Machine learning systems had become increasingly

present in today’s world in all kinds of domains, rang-

ing from object detection, prediction of weather, play-

ing games like chess or go (Silver et al., 2017b,a;

Mnih et al., 2013) prediction of a disease spread pat-

tern or even to self-driven cars. As stated by Adadi

and Berrada (2018), AI has already become ubiqui-

tous, and we have become accustomed to AI making

decisions for us in our daily life, from product and

movie recommendations on Netﬂix and Amazon to

friend suggestions on Facebook and tailored adver-

tisements on Google search result pages.

https://orcid.org/0000-0002-3094-5991

https://orcid.org/0009-0005-8179-1987

https://orcid.org/0000-0002-0940-9641

https://orcid.org/0000-0001-7884-2353

https://orcid.org/0000-0001-7552-6144

https://orcid.org/0000-0003-1916-8893

https://orcid.org/0000-0002-6334-0120

Disease spread prediction systems have gained

signiﬁcant attention in recent years, particularly in

the context of the COVID-19 pandemic, as they of-

fer valuable insights for controlling the spread of in-

fectious diseases and allocating resources for research

and development. Machine learning methods, such as

artiﬁcial neural networks and long short-term mem-

ory (LSTM) networks, have been increasingly em-

ployed in time series forecasting and predicting the

number of daily cases, deaths, and recovered cases

of diseases like COVID-19. These models can help

public health professionals and policymakers make

informed decisions based on the analysis and predic-

tions provided by the models, ultimately contributing

to the effective control of disease outbreaks.

The implementation of intelligent systems plays a

pivotal role in the control and prevention of animal

diseases, signiﬁcantly impacting public health. This

encompasses a range of strategic interventions, in-

cluding animal vaccination, vector control, stringent

hygiene protocols, and disease surveillance in ani-

Ebling, D., Machado, F., Descovi, G., Cardenas, N., Machado, G., Maran, V. and Machado, A.

A Distributed Processing Architecture for Disease Spread Analysis in the PDSA-RS Platform.

DOI: 10.5220/0012742700003690

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 26th International Conference on Enterprise Information Systems (ICEIS 2024) - Volume 2, pages 313-320

ISBN: 978-989-758-692-7; ISSN: 2184-4992

313

mals. The globalization and extensive trade in ani-

mals and their products have heightened the global

spread of zoonotic diseases, underscoring the urgent

need for robust prevention and control strategies to

avert potential epidemics. Moreover, the operational

dynamics within farms introduce inherent risks for

disease transmission. Factors such as the segrega-

tion of animal groups, human and vehicular trafﬁc

patterns, as well as the spatial layout of farms, can

signiﬁcantly inﬂuence the potential transmission of

diseases. For instance, the proximity of animal lots

can facilitate disease spread, while effective isolation

measures can help contain outbreaks. Human and ve-

hicular movement on farms can introduce pathogens,

emphasizing the importance of biosecurity measures.

Additionally, the spatial organization of farms can im-

pact pathogen dissemination, highlighting the need

for tailored control measures to mitigate disease trans-

mission risks effectively (Galvis et al., 2022).

In this area, Cardenas et al. (2021) proposed an

intelligent system that uses real data to simulate the

spread of diseases in animals (e.g., cattle, swine, and

small ruminants) and allows researchers to test differ-

ent actions to control these diseases. When the anal-

ysis is complete, the system makes the model avail-

able on a user interface for further action. This sys-

tem addresses the challenges of controlling the spread

of infectious diseases and provides valuable insights

for public health interventions. The use of Docker

containers for packaging the machine learning models

allows for easy deployment and scalability, enabling

rapid response and widespread intervention in situa-

tions where rapid deployment is necessary, such as

disease outbreaks. The development of disease spread

prediction systems using machine learning methods

offers a promising approach for controlling the spread

of infectious diseases and allocating resources for re-

search and development. By understanding the archi-

tecture of deployment of machine learning models in

this context, we can better control the spread of dis-

eases and mitigate their impact on society.

With that in mind, the objective of this work is

to build a tool to help veterinaries uses a stochas-

tic model thought a friendly interface that helps them

in critical decision-making situations and works on a

scalable way.

The disease control system serves as a valuable

tool; however, its current design caters primarily to

users with programming expertise who can navigate

the installation of dependencies and interact with it

through a programming interface. This is evident in

its distribution as a downloadable library, which ne-

cessitates a certain level of technical proﬁciency. Fur-

thermore, the system’s analysis demands substantial

computational resources, with tests indicating a re-

quirement of over 4GB of RAM for a single analysis

to run effectively. This high technical barrier poses a

signiﬁcant challenge for widespread adoption, partic-

ularly among veterinary professionals who may lack

programming skills. Accessing and utilizing the sys-

tem not only demands knowledge of livestock dis-

eases and epidemiology but also proﬁciency in the

system’s programming language. Addressing this is-

sue was the primary impetus behind this research en-

deavor: to enhance accessibility for epidemiologists

by enabling them to input their data efﬁciently into

the existing tool developed by Descovi et al. (2021).

The present paper is structured as follows. Section

2, presents the background, giving a brief introduction

about the stochastic model developed by Cardenas

et al. (2021) that models the disease spread and con-

trol system, and presents the problem that this work

will pursue. Section 3 describes the proposed sys-

tem to integrate the model and provide a solution to

the problem described. Section 4 describes our eval-

uation, a evaluation scenario to validate the system.

Section 5 outlines the conclusions of this work.

2 BACKGROUND AND

MOTIVATION

This section provides an overview of the Disease

Spread and Control System utilized in this study,

highlighting the existing challenges that will be ad-

dressed within this research.

2.1 Disease Spread and Control System

(Cardenas et al., 2022a, 2021, 2022b) introduced a

model designed to demonstrate the potential for the

spread of infectious animals within livestock popula-

tions, considering transmission via animal movement.

This model incorporates simulations of various con-

trol actions, such as preventing farm entry or exit of

infected animals (e.g., culling, isolation of animals),

enhancing hygienic practices, or implementing vacci-

nation programs.

This system offers an opportunity for users to

forecast the impact of infectious outbreaks and de-

velop proactive strategies to mitigate risks. By lever-

aging computational modeling, the system provides

insights into the dynamics of disease propagation un-

der diverse scenarios, enabling stakeholders to make

informed decisions regarding prevention, interven-

tion, and response efforts. The core functionality of

the system lies in simulating the spread of infectious

agents among livestock populations while accounting

ICEIS 2024 - 26th International Conference on Enterprise Information Systems

314

for factors like animal movement patterns, biosecurity

measures, and environmental conditions. Users can

input real-world data related to animal demographics,

geographic locations, and historical disease occur-

rences, allowing them to generate customized models

tailored to speciﬁc regional needs.

Once a model is created, users can explore various

what if scenarios by testing alternative control action

plans against the simulated outcomes. For instance,

they might examine the consequences of imposing

quarantine zones around affected farms, assessing the

efﬁcacy of targeted vaccination campaigns, or analyz-

ing the cost-beneﬁt tradeoffs associated with differ-

ent control options. By providing a comprehensive

decision support framework, the system empowers

policymakers, veterinary practitioners, and producers

alike to anticipate and respond effectively to emerg-

ing threats posed by infectious diseases in livestock

populations. This work focuses on using this model

to provide a tool for veterinaries to take actions and

explore scenarios, helping them in critical decision-

making situations.

2.2 Animal Movement Representation

The system uses Social Network Analysis (SNA)

methods to characterize animal trade patterns, and the

between-farm total of animals moved is represented

in the system as a directed graph, where each farm

is represented as a node, and the movements among

farms are represented as edges. Each edge connects

a speciﬁc node origin to a speciﬁc node destination,

also maintaining the type and number of animals be-

ing moved. An example of graph movement repre-

sentation is shown in Figure 1.

Figure 1: Between-farm animal movement graph.

The system maintains an event dataset with data

on the origin, destination, type, and number of ani-

mals, which is used to construct the movement graph.

The between-farm movements among farms of dif-

ferent species, therefore considering a real multi-host

contact network of movement data collected.

2.3 Disease Spread Dynamics

The system applies a stochastic simulation algorithm

(SSA) to simulate the disease spreading, as well as,

vital dynamics (birth and deaths) of animals inside

each farm. The system incorporates within farm and

between farm dynamics through a susceptible infec-

tious model using the temporal animal movement data

explicitly with a higher effective contact rate to en-

sure an efﬁcient disease transmission over the simula-

tions. Figure 2 illustrates the transition of states dur-

ing within-farm and between-farm dynamics on the

model.

Figure 2: Disease spreading simulation (Cardenas et al.,

2022a).

2.4 Control Action Zones

After an initial covert proliferation wherein animals

on select farms contracted the disease, expedient con-

trol and conﬁnement of the pathogen are paramount

for eliminating the infection and fostering recupera-

tion. Measures such as quarantining and regulating

animal movements serve to shield animal health, im-

peding the transmission of illness to uncontaminated

populaces (Roth, 2007). These actions are executed

within designated control regions, whose boundaries

are determined via user parameters. However, despite

the effectiveness of these measures, there remains a

pressing need to streamline and simplify the applica-

tion process for veterinary practitioners. Currently,

the implementation of these controls requires special-

ized knowledge and technical competency, limiting

the reach and utility of these measures among the

broader veterinary community. To bridge this gap,

this research aims to develop a user-friendly platform

that seamlessly integrates the Disease Spread and

Control System, allowing veterinary professionals to

effortlessly apply these protective measures with min-

A Distributed Processing Architecture for Disease Spread Analysis in the PDSA-RS Platform

315

imal technical expertise required.

2.5 Vaccination

Within the system, the vaccination process is simu-

lated throughout the course of disease spread. Upon

vaccination, animals transition from the Susceptible-

Exposed-Infectious-Recovered (SEIR) compartments

to a distinct Vaccinated (V) compartment. This segre-

gation allows us to precisely track the impact of vac-

cination on disease spread within the population.

The transition of animals into the V compartment

is inﬂuenced by two key factors: the efﬁcacy of the

vaccine and a user-deﬁned daily conversion rate for

control actions. The efﬁcacy of the vaccine deter-

mines its ability to confer immunity, affecting the pro-

portion of vaccinated animals protected from infec-

tion. Meanwhile, the daily conversion rate dictates the

pace at which susceptible individuals are vaccinated,

thus inﬂuencing the rate of accumulation within the V

compartment over time

3 RELATED WORKS

In this section, we present a brief overview of two

related works that have similar goals to our research,

but it end fail to provide a suitable solution for our use

case.

The Australian Animal Disease Spread

Model (Bradhurst et al., 2015), AADIS, is a de-

cision support tool that assists in the formulation

of policies and response strategies for emergency

animal diseases. The model integrates mathematical,

agent-based, network, and cellular automata mod-

elling approaches to simulate the incursion, detection,

surveillance, control, and proof-of-freedom of emer-

gency animal diseases. AADIS is designed to provide

valuable insights into the spread and management

of animal diseases, aiding in the development of

effective response strategies.

The Animal Disease Spread Model (Schoenbaum

et al., 2024), (ADSM), is a stochastic, spatially ex-

plicit compartmental model that simulates the spread

of highly contagious animal diseases between herds.

Developed by the USDA-APHIS-VS-CEAH, ADSM

is designed to evaluate different control strategies,

provide recommendations on resource allocation, and

estimate the economic impacts of disease control op-

tions. The model is based on the North American Dis-

ease Spread Model (NAADSM) and has been used to

simulate foot-and-mouth disease (FMD) outbreaks in

the United States, focusing on stamping-out, slaugh-

ter, and vaccination control strategies. ADSM is built

in Django and Python and presented as a web applica-

tion, allowing users to interact with the model through

a user-friendly interface. The model uses multithread-

ing to maximize system utilization and provides de-

tailed outputs and supplemental outputs for further

analysis.

The two works presented in this sections provide

a model for a dieases spread, and a UI for users. The

main problems with both works is that they provide a

application to be run in the machine of the user, be-

ing a desktop focused application, wheere a end-user

not only would need to setup their environment to be

able to run the application, have a good knowledge

of the disease and the model to be able to use the ap-

plication, and also have a good enought machine to

run the simulation. This issues makes the application

not user friendly and insuitable for our use case. With

that problem also comes the fact that the models are

closed source, so we can’t use the code to make our

own changes and provide a application for the user,

making both works not suitable for our use.

4 ARCHITECTURE APPROACH

The proposed solution for analyzing the spread of dis-

ease in a farm scenario is a web system that com-

bines a user interface (UI), an application program-

ming Interface (API), and a Python machine learning

model to provide a comprehensive and user-friendly

platform for tracking and predicting the spread of dis-

eases. The system consists of 3 main components,

a client API that will integrate with the system, the

model’s API which will dispatch, manage and scale

an instance of the model, and the model itself.

The client API will be the system that holds the

data, and want to integrate the model. The client API

has the data about premises, movement and events,

and can use it to query the model’s API to create and

view analysis. Besides having the data to analyze,

the only required implementation on this component,

is that it needs to know how to communicate with a

RESTful(Ehsan et al., 2022) API.

The stochastic model is originally implemented as

a library Cardenas et al. (2021), so the next compo-

nent is a thin wrapper around the model that provides

two important capabilities that allow it to be managed

by the API: A ability to be run as a Docker(Merkel,

2014) container, and a JSON-based based protocol for

communication via STDIO. The ability to be run as

a Docker container is fundamental, where that will

be used to scale the model the instances, and en-

able the API to control its resources, and the JSON-

based based protocol that allows the API to moni-

ICEIS 2024 - 26th International Conference on Enterprise Information Systems

316

tor communicate with it to get real-time data about

the progress of an analysis. The choice of using the

STDIO transport, instead of TCP or other common

transport mechanism was heavily inspired by the Mi-

crosoft’s Language Server Protocol (LSP) implemen-

tations, where the most of the implementations use

the STDIO to communicate with an editor using a

JSON-based based protocol to send information about

a project. With this, we get a RESTful APIs commu-

nicating with JSON protocol and the model commu-

nicating also with a JSON protocol but via STDIO

with the Model’s API, this is illustrated in Figure 3,

where it shows the type of communication between

the components.

The model’s API is the entrypoint to access the

stochastic model capabilities, is responsible to spawn,

schedule, and managing the request for analysis us-

ing the model. It’s implemented as a RESTful API

that communicates with the client API for data man-

agement to provide real-time updates on the dis-

ease spread predictions and the impact of intervention

strategies. Its main concern is to bridge client APIs

and the disease spread model. It does that by manag-

ing a pool of workers for the model, where when an

analysis request is made, the job is dispatched to this

pool, where each work is run as a Docker container.

The API spawns and monitor this container STDOUT

stream, parsing its log messages containing JSON in-

formation, with this the API gets real-time data about

the progress of the model. Another important role of

this API is to aggregate the result of multiple workers.

Some analysis requests will need to dispatch hundreds

of instances of the model, and the result of the con-

tainer needs to be aggregated in representation of the

mean result of the all model instances. THe API does

that by persisting the data about the running models

in a relational database, where later will be queried to

provide the results for the client APIs. With this, it

provides a scalable model to run the stochastic mod-

els, where the work can be dispatched into a Docker

cluster to run.

With all components together, an analysis can be

made starting from the client API, where it will send

data to be analyzed to the model’s API, together with

the parameters to run the model. The model’s API

will spawn N workers to run this analysis, monitoring

and aggregating its output. In the meantime, the client

can query the API to get real-time progress about the

analysis, and when it’s done, will have full access to

the output of the model. This ﬂow of the process is

illustrated in Figured 4

In conclusion, the proposed solution combines

multiple technologies, to provide a scalable solution

to running a stochastic model. This system offers

valuable insights for controlling the spread of dis-

eases and allocating resources for research and devel-

opment.

5 CASE STUDY SCENARIO

To evaluate the feasibility of the approach, and im-

plementation developed, a case study was developed,

This case study aims to demonstrate the use of our

application in a scenario where there’s a suspect of

a disease outbreak on premises. This application has

the main goal of helping veterinaries combat and con-

trol outbreaks proactively.

This ﬁctional scenario is based in the everyday

work of state veterinaries in Brazil, that aims to as-

sess the effective of the application in helping the vet-

erinary control and identify other farms that could be

affected.

For the case study, the following scenario was

considered for describing the approach supported by

the developed application. Imagine ”John”, John

is ofﬁcial Veterinary of the State, in Brazil. As a

State Veterinary, ones of this many tasks is to audit

premises and certify that they are clean from common

diseases that could be detrimental to the livestock pro-

duction ecosystem.

One day, while conducting routine audits, John is

notiﬁed that there is simptons of a disease in Arnold’s

premises. Suspicious symptoms in some animals and

unusual mortality rates raise concerns. Recognizing

the urgency of the situation, John swiftly takes sam-

ples and sends them for testing to conﬁrm the pres-

ence of a contagious disease.

Upon receiving the test results, John logs into the

PDSA-RS’s (Descovi et al., 2021) system to use the

disease control module to help him plan his actions.

The system provides a user-friendly interface where

he inputs the conﬁrmed case and the relevant details.

Leveraging the application’s advanced algorithms, it

quickly analyzes the data, considering factors such

as geographical proximity, animal movement records,

and environmental conditions.

In a matter of minutes, the application generates a

comprehensive report outlining the potential risk and

identiﬁes other premises at high risk of being affected

by the outbreak. The predictive modeling algorithms

take into account various factors, including animal

transportation networks, wind patterns, and historical

disease spread data. This gives John various infor-

mation, showing how this disease will spread infect

premises around even from other species, this is illus-

trated in Figure 5.

John is presented with a map highlighting the

A Distributed Processing Architecture for Disease Spread Analysis in the PDSA-RS Platform

317

Figure 3: System architecture diagram.

Figure 4: Communication ﬂow between the client and the model’s API to create a analysis.

Figure 5: Application showing a line plot illustrating the

amount of infected premises over time.

farms most susceptible to the contagion, as shown

in Figure 6. The system also provides an estimated

timeline for potential outbreaks on these identiﬁed

premises, allowing John to prioritize and plan emer-

gency response measures effectively.

Figure 6: Application showing a disease spread map.

To plan his actions, John, uses the control actions

feature of the application, where it lets him apply ac-

tions to a set of premises and see what’s the impact

of the actions to control the outbreak of that diseases.

This becomes an iterative process, where the vet will

update the system with the current state of the spread

of the diseases, use the system to help plan the next

actions, execute the actions and restart the process

until the outbreak is contained. This process is illus-

trated on Figure 7.

Equipped with this vital information, John ini-

tiates immediate communication with the identiﬁed

farms, informing them of the potential threat and ad-

vising precautionary measures. The system’s real-

time capabilities enable John to monitor the situation

as it unfolds, receiving updates on the progression of

the outbreak and making informed decisions to con-

tain its spread.

Under the hood, everytime that the user create a

new analysis or updates one with the current infected

data, the system undergoes the following process to

produce a report:

• PDSA-RS requests a new analysis, by sending the

premises, movement and events data to the model

api.

• The model’s Scheduler API will create one or

more docker containers, running the model in-

stances using a pool of workers algorithm. Here

the API will spawn a new job and propagate the

data to them.

ICEIS 2024 - 26th International Conference on Enterprise Information Systems

318

Figure 7: Process to control an outbreak of a disease.

• The API will aggregate the output of all model

instances and its store the ﬁnal data.

• While the Scheduler is doing its works, the

PDSA-RS will query its status, and when its ready

will request the processed result of the analysis

In this way, the application proves instrumen-

tal in empowering veterinarians like John to respond

rapidly and strategically to disease outbreaks, mini-

mizing the impact on livestock and safeguarding the

overall health of the agricultural ecosystem. The

scenario showcases the practicality and effectiveness

of the developed application in a real-world setting,

demonstrating its potential to revolutionize disease

control efforts in the livestock industry.

6 CONCLUSIONS

This case study was successful, on this work we built

a tool that helps veterinaries make critical decisions

and explore possibilities in advents of an outbreak of

a diseased, allowing them to make conﬁdent decisions

to control and to contain outbreaks of diseases in the

livestock production industry.

Based on the results of the case study scenario,

we accommodate the users needs on the tool, and of a

test made on 23rd, October 2023, on a workshop with

the state veterinaries from the state of Rio Grande do

Sul.For future work it will be important to focus on

reducing even more the costs, making the application

more scalable for general users. Another improve-

ment for future work would be the UI/UX of the ap-

plication, while attending the user needs we noticed

that there’s room for improvement for an easier to un-

derstand UI.

ACKNOWLEDGEMENTS

This research is supported by FUNDESA, project Ap-

plication of Machine Learning Techniques to Predict

the Prevalence of Diseases in the Processes of Certi-

ﬁed Swine Breeding Fars and Monthly Pige Epidemi-

ological Sheet (UFSM/057438) and Research and De-

velopment of Innovative Technologies focused on

Agribusiness (UFSM/051568) and (UFSM/060642)

A Distributed Processing Architecture for Disease Spread Analysis in the PDSA-RS Platform

319

Use of artiﬁcial intelligence in the systematization

of hygienic sanitary certiﬁcation processes for ship-

ments and accreditation of legal origin of ﬁsh. The

research by Vincius Maran is partially supported by

CNPq grant 306356/2020-1 (DT-2).

REFERENCES

Adadi, A. and Berrada, M. (2018). Peeking inside the black-

box: A survey on explainable artiﬁcial intelligence (xai).

IEEE Access, 6:52138–52160.

Bradhurst, R. A., Roche, S. E., East, I. J., Kwan, P., and Gar-

ner, M. G. (2015). A hybrid modeling approach to sim-

ulating foot-and-mouth disease outbreaks in australian

livestock. Frontiers in Environmental Science, 3:17.

Cardenas, N. C., Lopes, F. P., and Machado, G.

(2022a). Modeling foot-and-mouth disease dissemina-

tion in brazil and evaluating the effectiveness of control

measures. bioRxiv, pages 2022–06.

Cardenas, N. C., Pozo, P., Lopes, F. P. N., Grisi-Filho, J. H.,

and Alvarez, J. (2021). Use of network analysis and

spread models to target control actions for bovine tuber-

culosis in a state from brazil. Microorganisms, 9(2):227.

Cardenas, N. C., Sykes, A. L., Lopes, F. P., and Machado,

G. (2022b). Multiple species animal movements: net-

work properties, disease dynamics and the impact of tar-

geted control actions. Veterinary Research, 53(1):14.

Descovi, G., Maran, V., Ebling, D., and Machado, A.

(2021). Towards a blockchain architecture for animal

sanitary control. In ICEIS (1), pages 305–312.

Ehsan, A., Abuhaliqa, M. A. M., Catal, C., and Mishra, D.

(2022). Restful api testing methodologies: Rationale,

challenges, and solution directions. Applied Sciences,

12(9):4369.

Galvis, J. A., Corzo, C. A., Prada, J. M., and Machado,

G. (2022). Modeling between-farm transmission dynam-

ics of porcine epidemic diarrhea virus: characterizing

the dominant transmission routes. Preventive Veterinary

Medicine, 208:105759.

Merkel, D. (2014). Docker: lightweight linux containers for

consistent development and deployment. Linux journal,

2014(239):2.

Mnih, V., Kavukcuoglu, K., Silver, D., Graves, A.,

Antonoglou, I., Wierstra, D., and Riedmiller, M. (2013).

Playing atari with deep reinforcement learning. arXiv

preprint arXiv:1312.5602.

Roth, J. A. (2007). Animal disease information and preven-

tion materials developed by the center for food security

and public health. Iowa State University Animal Industry

Report, 4(1).

Schoenbaum, M., Assefa, D., Holmstrom, L., Delgado, A.,

Seaman, J., Collins, F., and Springs, C. C. (2024). Sim-

ulating foot-and-mouth disease in the united states using

the animal disease spread model.

Silver, D., Hubert, T., Schrittwieser, J., Antonoglou, I., Lai,

M., Guez, A., Lanctot, M., Sifre, L., Kumaran, D., Grae-

pel, T., Lillicrap, T., Simonyan, K., and Hassabis, D.

(2017a). Mastering chess and shogi by self-play with

a general reinforcement learning algorithm.

Silver, D., Schrittwieser, J., Simonyan, K., Antonoglou, I.,

Huang, A., Guez, A., Hubert, T., Baker, L., Lai, M.,

Bolton, A., et al. (2017b). Mastering the game of go

without human knowledge. nature, 550(7676):354–359.

ICEIS 2024 - 26th International Conference on Enterprise Information Systems

320