| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| 7.39 MB | Adobe PDF |
Authors
Advisor(s)
Abstract(s)
A utilização de tecnologia tem evoluído de forma exponencial e também a necessidade de análise de dados para identificar padrões, obter métricas e descobrir tendências. Há cada vez mais dados disponíveis sobre processos de negócios e a mineração de processos visa preencher a lacuna que se verifica entre modelos de processos tradicionais e técnicas de análises de dados. É usada para extrair conhecimento através de registos de eventos. As técnicas de mineração de processos permitem descobrir novos modelos de processos, monitorizar e melhorar os processos já existentes. É na área da saúde que se enquadra o estudo de caso apresentado. Esta tem-se tornado um dos grandes focos de análise de processos de negócios, devido à crescente urgência em descobrir padrões, eliminar ineficiências de modo a melhorar a qualidade dos serviços, e, simultaneamente, reduzir custos financeiros e temporais. Devido à diversidade, dinamismo e complexidade dos processos médicos, torna-se difícil analisar os dados recolhidos e encontrar padrões. Grande parte dos dados recolhidos nos hospitais advêm do diagnóstico e tratamento dos pacientes. A aplicabilidade da mineração de processos na área da saúde auxilia à análise e monitorização dos percursos realizados por cada paciente, bem como, à deteção de anomalias, desvios e estrangulamentos nos processos existentes. Esta dissertação apresenta uma abordagem capaz de determinar processos de negócio relacionados com tratamentos clínicos obtidos através de registos de eventos armazenados na base de dados Medical Information Mart for Intensive Care III. Para atingir os objetivos propostos foi desenvolvido uma abordagem dividida por etapas utilizando a ferramenta ProM. Numa primeira fase os dados foram extraídos e mapeados para serem importados para a ferramenta ProM, o que realizado numa segunda fase, com algumas atividades de pré-processamento. Após importação, os dados foram transformados utilizando plugins específicos, e de seguida, aplicaram-se algoritmos de descoberta de processo. O resultado final são modelos de processo em notação business process model and notation ou rede de Petri. Por fim, os modelos de processo gerados foram processados e avaliados de acordo com métricas definidas.
Technology as a whole has exponentially evolved throughout the last few decades. There is also a clear need for data analysis to indentify patterns, obtain metrics and discover tendencies. Everyday, the number of data available about business processes is increasing. For this reason, the art of process mining aims to fil the gap between traditional process models and data analysis techniques. Therefore, it is used to extract knoweldge from event rigistries. Process mining techniques allow the discovery of new process models, monitoring and improvement of existent processes. Health care has been one of the most important areas of activity today where there is a clear need for pattern discovery which permit the improvement of the overall quality of services and, at the same time, drastically reduce temporal and financial costs. Moreover, due to the highly diversified, dynamic and complex medical processes, it is substantially difficult to analyse collected data and find patterns in them. Almost the entirety of collected data in health centers such as hospitals come from the diagnosis and treatment of patients. The applicability of process mining in health care greatly helps the analysis and motorization of the overall paths taken by each patient in addition to the detection of anomalies, deviations and bottlenecks in existing processes. This dissertation presents a possible approach capable of determining medical processes obtained from multiple event logs, which are stored in the Medical Information Mart for Intensive Care III database. In order to achieve this dissertation objectives, a multiple phased approach was developed by using the ProM tool. In a first stage, the data provided by the medical center was extracted and mapped. At a second phase, the data was transformed using multiple specific plugins in addition to the application of knowledge process discovery algorithms. The final result is several process models in the Business Process Model and Notation and Petri Net notations. Finally, the generated process models were processed and evaluated according to multiple defined metrics.
Technology as a whole has exponentially evolved throughout the last few decades. There is also a clear need for data analysis to indentify patterns, obtain metrics and discover tendencies. Everyday, the number of data available about business processes is increasing. For this reason, the art of process mining aims to fil the gap between traditional process models and data analysis techniques. Therefore, it is used to extract knoweldge from event rigistries. Process mining techniques allow the discovery of new process models, monitoring and improvement of existent processes. Health care has been one of the most important areas of activity today where there is a clear need for pattern discovery which permit the improvement of the overall quality of services and, at the same time, drastically reduce temporal and financial costs. Moreover, due to the highly diversified, dynamic and complex medical processes, it is substantially difficult to analyse collected data and find patterns in them. Almost the entirety of collected data in health centers such as hospitals come from the diagnosis and treatment of patients. The applicability of process mining in health care greatly helps the analysis and motorization of the overall paths taken by each patient in addition to the detection of anomalies, deviations and bottlenecks in existing processes. This dissertation presents a possible approach capable of determining medical processes obtained from multiple event logs, which are stored in the Medical Information Mart for Intensive Care III database. In order to achieve this dissertation objectives, a multiple phased approach was developed by using the ProM tool. In a first stage, the data provided by the medical center was extracted and mapped. At a second phase, the data was transformed using multiple specific plugins in addition to the application of knowledge process discovery algorithms. The final result is several process models in the Business Process Model and Notation and Petri Net notations. Finally, the generated process models were processed and evaluated according to multiple defined metrics.
Description
Keywords
Business process model and notation Descoberta de processos Mineração de processos Processos de negócio Rede de Petri Registo de eventos