Showing 2 results for Neural Network
Asieh Khosravanian, Saeed Ayat,
Volume 18, Issue 2 (2-2019)
Abstract
Backgrounds: Early detection of diabetes is critical to avoid complications and damage caused by this disease. The purpose of this paper is designing an intelligent system for Diabetes prediction (healthy or patient) by using regression method based on Multilayer Perceptron Neural Network.
Methods: In this descriptive-analytic study, an intelligent system is designed to classification diabetes patients. The system is simulated by MATLAB software 2015 (8.5.0.197613). In this study, used PID dataset in UCI Machine Learning Repository. The dataset is contained 768 records from Indian women and 8 diagnostic factors for Diabetes.
Results: The data were then divided randomly in 20 groups for training and testing, after preprocessing. 90% of the data is used for training phase and 10% for the test phase. The results obtained based on sensitivity, specificity, accuracy and precision were 0.4815, 0.9804, 0.8077 and 0.9286, respectively.
Conclusion: The obtained results, showed superiority of designed intelligent system to classify individuals (healthy and patient) in comparison with other methods implemented on this dataset. Using MLP- Regression has increased the accuracy of the proposed system.
Abolfazl Kazemi, Hamid Bahador,
Volume 21, Issue 3 (9-2021)
Abstract
Background: Today, in most hospitals in Iran, there is an extensive database of patient characteristics that includes a large amount of information related to medical, family and medical records. Finding a knowledge model of this information can help to predict the performance of the medical system and improve educational processes.
Methods: Data mining techniques are analytical tools that are used to extract meaningful knowledge from a large data set. In this study, the information of 500 people referred to Shahid Bolandian Health Center in Qazvin has been used. In this research, a predicted model has been performed using decision tree data mining methods and neural network and Bayesian network.
Results: The decision tree model has the highest accuracy and the Bayesian network has the lowest accuracy in diagnosing diabetic patients, and consequently the decision tree has the least error and the Bayesian network has the highest error. The decision tree model with 95.68% had the highest accuracy in prediction.
Conclusion: Fat has the greatest effect in predicting diabetes and gender has the least effect in predicting diabetes. Based on the decision tree analysis, the rules obtained among the stated characteristics of age and sugar variables have the greatest effect in predicting the occurrence of diabetes (according to software analysis) and by creating a proper diet can prevent this disease Prevented.
