Modeling Breast Cancer using data mining methods

Introduction: Breast cancer is the most common form of cancer in women. Breast cancer detection is considered as one of the most important issues in medical science. Diagnosis of benign or malignant type of cancer reduces costs and also is important in deciding about the treatment strategy. The aim of this study was to provide data mining based models that have the predictability of breast cancer detection

Methods: This study was descriptive-analytic. Its database included 683 independent records containing nine clinical variables in the UCI machine learning. Multilayer Perceptron artificial neural network, Bayesian Neural Network and LVQ neural network were used for classification of breast cancer to benign and malignant types. In this study, 80 percent of data were used for network training and 20 percent were used for testing

Results: After pre-processing the data, different neural networks with different architectures were used to detect breast cancer. In the best condition, we could predict benign or malignant cancer in the MLP neural networks, LVQ and Bayesian Neural Networks with an average of ten tests with an accuracy of 97.5 percent and 97.6percent and 98.3 percent respectively. Our investigations showed that Bayesian neural network had a better performance

Conclusion: Breast cancer is one of the most common cancers among women. Early diagnosis of disease reduces healthcare costs and increases patient survival chance. In this study, using data mining techniques in diagnosis, the researchers were able to use Bayesian neural network to achieve high accuracy in diagnosis

Diagnosis of Coronary Heart disease using mixture of experts method

Introduction: Coronary Artery Disease [CAD] is one of the most common heart diseases and the main cause of mortality in men and women. This study aimed to predict the disease status using Neural Network compound [mixture of experts]

Methods: The present study was a diagnostic study conducted on 200 patients referred to a heart specialty center in Torbat-e-Heydarieh. Patients' files contained their demographic information including13 risk factors. A model for predicting CAD based on multilayer perceptron neural network and mixture of experts was produced

Results: First, we used a neural network of multilayer perceptron with Propagation algorithm by different architectures. The best architecture could predict closed coronary artery with the accuracy of 71.7 percent. Then, by increasing the number of neural networks and training process, results were combined. Mixture of experts by liner method [majority voting] and nonlinear method [gating network] was applied and the accuracy rates of 75.8 percent and 78.3 percent were respectively obtained

Conclusion: Angiography is an invasive diagnostic procedure with risk factors such as stroke and heart attack. Therefore, non-invasive methods should be used for the diagnosis of CAD. In this study, with increasing the number of learners and their nonlinear mixture, the accuracy of diagnosis was increased