Loading...
1 results
Search Results
Now showing 1 - 1 of 1
- Performance Evaluation of Smart Decision Support Systems on HealthcarePublication . Moreira, Mario Wedney de Lima; Rodrigues, Joel José Puga CoelhoMedical activity requires responsibility not only from clinical knowledge and skill but also on the management of an enormous amount of information related to patient care. It is through proper treatment of information that experts can consistently build a healthy wellness policy. The primary objective for the development of decision support systems (DSSs) is to provide information to specialists when and where they are needed. These systems provide information, models, and data manipulation tools to help experts make better decisions in a variety of situations. Most of the challenges that smart DSSs face come from the great difficulty of dealing with large volumes of information, which is continuously generated by the most diverse types of devices and equipment, requiring high computational resources. This situation makes this type of system susceptible to not recovering information quickly for the decision making. As a result of this adversity, the information quality and the provision of an infrastructure capable of promoting the integration and articulation among different health information systems (HIS) become promising research topics in the field of electronic health (e-health) and that, for this same reason, are addressed in this research. The work described in this thesis is motivated by the need to propose novel approaches to deal with problems inherent to the acquisition, cleaning, integration, and aggregation of data obtained from different sources in e-health environments, as well as their analysis. To ensure the success of data integration and analysis in e-health environments, it is essential that machine-learning (ML) algorithms ensure system reliability. However, in this type of environment, it is not possible to guarantee a reliable scenario. This scenario makes intelligent SAD susceptible to predictive failures, which severely compromise overall system performance. On the other hand, systems can have their performance compromised due to the overload of information they can support. To solve some of these problems, this thesis presents several proposals and studies on the impact of ML algorithms in the monitoring and management of hypertensive disorders related to pregnancy of risk. The primary goals of the proposals presented in this thesis are to improve the overall performance of health information systems. In particular, ML-based methods are exploited to improve the prediction accuracy and optimize the use of monitoring device resources. It was demonstrated that the use of this type of strategy and methodology contributes to a significant increase in the performance of smart DSSs, not only concerning precision but also in the computational cost reduction used in the classification process. The observed results seek to contribute to the advance of state of the art in methods and strategies based on AI that aim to surpass some challenges that emerge from the integration and performance of the smart DSSs. With the use of algorithms based on AI, it is possible to quickly and automatically analyze a larger volume of complex data and focus on more accurate results, providing high-value predictions for a better decision making in real time and without human intervention.