An Artificial Intelligence Hybrid Model Based on Neural Networks in Decision Support Systems with Medical Applications
The data for chemical injuries and patients are insufficient due to very low usage of chemical weapons in the world and the low number of people injured by using these weapons. Also, considering the importance of predicting cardiac-pulmonary test results on chemical injuries' data and the high durations and costs of conducting these tests as well as the possibility of data loss at different stages (e.g., user error in recording the results), this type of data is valuable. In this study, missing data are addressed, isolated, and K-means method is used to fill the missing values. Then, the data mining models based on neural network are applied for forecasting. In the next step, the neural network is utilized to estimate the outputs of cardiopulmonary tests based on the inputs (Spirometer tests). Also, the optimal number of neurons in hidden layer is investigated at this stage and the results of errors are reviewed. Finally, this article presents a hybrid model of artificial intelligence considering data preprocessing method and neural network to design decision support systems for medical applications. This article enables physicians to study, diagnosis and treat the disease as soon as possible before conducting time-consuming experiments.
Keywords - Data Mining, Neural Networks, Data Preprocessing, Clustering, Medical Application