Abstract:

Artificial intelligence has significantly benefited the development of medical informatics and biomedicine by making some approaches available for application, ranging from rule-based expert systems and fuzzy logic to neural networks and evolutionary algorithms. Evolutionary techniques are well known for handling nonlinear restricted optimization problems. After a few generations, the population typically converges to an area around the global optimum, thanks to the exploratory power of evolution-based optimizers. Although the search space can be effectively reduced by this convergence, most of the current optimization techniques continue the search over the original space, losing a significant amount of time on searching unsuccessful regions. One of the most important unresolved problems in pattern recognition, feature selection is an NP-hard problem from the perspective of algorithm creation. We propose a novel evolutionary-incremental feature selection method in this study. The suggested method allows for the application of a standard evolutionary algorithm (EA), such as a genetic algorithm (GA).To make typical EAs compatible with solutions that can have a variable length, this framework suggests certain general adjustments. The solutions about the primary generations in this framework are concise. The length of solutions can then be steadily expanded through generations. The primary components of the suggested solutions are (a) regular recording and monitoring of the patient's vital sign measures taken at home and (b) management of the patient's electronic medical records (EMRs).It aids in the pursuit of superior patient monitoring at home and prompt emergency response

Description:

Artificial intelligence; feature selection; medical conditions; learning method; genetic algorithms