Breast cancer classification based on analysis of patients' DNA arrays

Abstract

Cancer, especially breast cancer, is one of the most common causes of death worldwide according to the World Health Organization. For this reason, extensive research efforts have been made in the field of accurate and early cancer diagnosis in order to increase the odds of cure. Microarray technology has proven effective among the tools available for cancer diagnosis. Microarray technology analyzes the expression level of thousands of genes simultaneously. Although the sheer number of features or genes in microarray data may appear useful, many of these features are irrelevant or redundant leading to deterioration in classification accuracy. To overcome this challenge, feature selection techniques are a mandatory preprocessing step before the classification process.Methods and Materials: Artificial Neural Network (ANN), Support Vector Machine (SVM), Decision Trees (DT), Random Forest algorithms are used along with the most common statistical method (logistic regression). Building prediction models using a large data set the nature of the categories, performance evaluation criteria are defined in advance.Results: After applying different algorithms to the selected data, the results showed that the decision tree (DT) achieved an accuracy of 89.1% thanks to its ability to deal with multi-dimensional data and reduce the overlap between features. The performance of the neural network (ANN) was improved to 90.2% when applying reinforcement learning techniques ( Boosting), while deep neural networks (DNN) achieved the highest accuracy of 92.5%.Conclusion: A comparative study of multiple breast cancer predictive models using a large data set and three classification methods gave us insight into the relative predictive power of different data mining. After analyzing the data, we came to this conclusion: This model can be easily integrated into electronic health records (EHR) systems within hospitals and clinics To improve early diagnosis of breast cancer. Medical teams can be trained to use the model effectively through customized training programs, reducing human error and speeding up the diagnostic process.