Pearson Correlation Based Linear Discriminative Feature Selection For Breast Cancer Detection With Microarray Data
Breast cancer detection is a challenging task in healthcare industries. Early detection and treatment of breast cancer avoids the patient from serious issues. Accurate breast cancer identification is carried out with the selection of relevant features from dataset. Recently several research works were developed to choose more relevant feature for detecting the breast cancer. Existing Nested Genetic Algorithm (Nested-GA) was designed to select relevant features from microarray dataset. However, breast cancer detection accuracy was not effectively improved. Besides, existing hybridized approach based on feature selection and feature extraction was developed to discover the cancer. However, the time complexity was higher. In order to overcome the above limitations, an effective feature selection technique is needed for breast cancer detection.
Therefore, Pearson Correlated Discriminant Function Analysis Based Feature Selection (PCDFA-FS) technique is developed for increasing the accuracy of breast cancer detection at an earlier stage. The objective of proposed PCDFA-FS technique is to discover the cancer gene expression data with higher accuracy and less time complexity. The above said aim is achieved with the performance of Discriminant Function Analysis in proposed PCDFA-FS technique. The number of gene expression data is taken as input in PCDFA-FS from given microarray dataset. To start with proposed PCDFA-FS technique, the number of classes is initialized (relevant or irrelevant). Then the correlation between features and objective (i.e., breast cancer detection) is determined with the help of computing Pearson Correlation Coefficient. According to the correlation measure, the discriminant vector projects the features into the relevant or irrelevant classes. From that the proposed PCDFA-FS technique increases the accuracy of breast cancer detection with minimum time consumption and false positive rate in proposed PCDFA-FS technique.