Main Article Content
Cancer is one of the slowly progressing diseases that exhibit symptoms only at the later stage of the disease. Cancer that is common among women is breast cancer and in recent years, the total number of women affected has elevated to a higher number across the globe. It is more prevalent in the western world than on the other side of the world due to varied food habits and stressful lifestyles. To understand the various factors that contribute to the development of the disease, to classify the disease as benign or malignant and for predicting the disease several machine learning models were employed. In a similar perspective, machine learning models can be also be utilized to identify or predict potential breast cancer drugs and classify them. This computational approach helps in reducing experimental costs that incur during the pre-clinal trials and enables to filter few potential drugs among millions of compounds available. The result relies on the type of feature set or attributes considered for the study. Prediction of the drug is determined based on the feature set that defines the physicochemical, lipophilicity, water-solubility, pharmacokinetics, and drug-likeness properties of the compound. In this paper, a new multiphase feature selection with pipelined methodology is proposed that enhances the prediction accuracy of the breast cancer drug. This study further investigates the significance of feature selection and its impact on the predicted result. Multilayer perceptron model obtained high accuracy of 94.7% compared to the other supervised machine learning models.