Sistema de predição de risco de quedas para área hospitalar utilizando machine learning
Carregando...
Citações na Scopus
Tipo de produção
Trabalho de Conclusão de Curso
Data
2021-12-09
Autores
Pereira, Isaías Cabral
Batista, Jonathan Fellipe Francisco
Sonoda, Matheus Salgueiro
Nacimento, Monica Marques
Braga, Vinícius Silva
Batista, Jonathan Fellipe Francisco
Sonoda, Matheus Salgueiro
Nacimento, Monica Marques
Braga, Vinícius Silva
Orientador
Tonidandel, Flavio
Periódico
Título da Revista
ISSN da Revista
Título de Volume
Citação
Texto completo (DOI)
Palavras-chave
queda idosos,predição,inteligência artificial,falls,elderly,risks,prediction
Resumo
O mundo vive um crescente momento de envelhecimento populacional. Vários fatores explicam tal fenômeno, como por exemplo a queda da fecundidade proveniente da diminuição do número de nascimentos, além do aumento da expectativa de vida da população em decorrência de melhorias no âmbito da saúde e na qualidade de vida. Por outro lado, o problema de quedas sofridas pela população idosa já é um assunto preocupante conhecido ao redor do mundo, uma vez que pequenas quedas podem gerar grandes transtornos e riscos à vida, saúde e integridade física desta camada da população. Com o agravante ainda de que no momento atual muitos idosos moram sozinhos e devido ao afastamento de seus familiares, os riscos se tornam ainda maiores. Devido ao aumento no número de quedas, diversos dispositivos já foram propostos para detectá-las e em seguida alertar aos contatos previamente cadastrados pelo usuário. Nesse sentido, estudar os indícios de uma possível queda é de suma importância, uma vez que ela pode trazer sérias consequências para a saúde da população, podendo até mesmo ser fatal. O objetivo deste trabalho consiste, portanto, no desenvolvimento de uma ferramenta de predição de risco de queda por meio da utilização dos algoritmos de machine learning árvore de decisão (DT) e rede neural, por meio dos softwares MATLAB® e WEKA. Isto posto, foram treinados dois modelos computacionais baseados nestes algoritmos, que puderam predizer o risco de quedas no idoso. Tais modelos foram embasados nas informações contidas num banco de dados previamente estabelecido e obtiveram taxas de assertividade, precisão e sensibilidade adequadas com possibilidades de melhorias para aplicação prática futura.
The world is experiencing a growing moment of population aging, several factors ex- plain such a phenomenon, such as the fall in fertility resulting from the decrease in the number of births, in addition to the increased life expectancy of the population due to improvements in health and quality of life. On the other hand, the problem of falls suffered by the elderly population is already a matter of concern known around the world, since small falls can cause major problems and risks to the life, health, and physical integrity of this population group. With the aggravating factor that nowadays many elderly people live alone and due to the distance from their families, the risks become even greater Due to the increase in the number of falls, several devices have already been proposed to detect them and then alert the contacts previously registered by the user. In this sense, studying the signs of a possible fall is of utmost importance, since it can bring serious consequences to the population’s health, and can even be fatal. The objective of this work is, therefore, to develop a tool to predict the risk of falling by using the machine learning algorithms decision tree (DT) and neural network, through the MATLAB® and WEKA. That said, two computational models were trained based on these algorithms, which were able to satisfactorily predict the risk of falls in the elderly. These models were based on information contained in a previously established database and obtained adequate assertiveness, precision and sensitivity rates, with possibilities of improvement for future practical application.
The world is experiencing a growing moment of population aging, several factors ex- plain such a phenomenon, such as the fall in fertility resulting from the decrease in the number of births, in addition to the increased life expectancy of the population due to improvements in health and quality of life. On the other hand, the problem of falls suffered by the elderly population is already a matter of concern known around the world, since small falls can cause major problems and risks to the life, health, and physical integrity of this population group. With the aggravating factor that nowadays many elderly people live alone and due to the distance from their families, the risks become even greater Due to the increase in the number of falls, several devices have already been proposed to detect them and then alert the contacts previously registered by the user. In this sense, studying the signs of a possible fall is of utmost importance, since it can bring serious consequences to the population’s health, and can even be fatal. The objective of this work is, therefore, to develop a tool to predict the risk of falling by using the machine learning algorithms decision tree (DT) and neural network, through the MATLAB® and WEKA. That said, two computational models were trained based on these algorithms, which were able to satisfactorily predict the risk of falls in the elderly. These models were based on information contained in a previously established database and obtained adequate assertiveness, precision and sensitivity rates, with possibilities of improvement for future practical application.