Breast cancer is one of the leading type of cancer impacting women globally, early diagnosis of such diseases helps in identifying prevalent symptoms in the early stages hence this study focuses on choosing best machine learning approaches to tackle this problem. In this study, machine learning approaches are presented to classify breast cancer sub-types based on high dimensional Gene Expression data. 151 samples with 54,676 gene activity features per sample have been obtained from Kaggle which are clustered into Basal, HER2-enriched, Luminal-A, Luminal-B, Cell Line, and Normal. The data set was rectified for class imbalance by the means of the Synthetic-Minority-Oversampling Technique (SMOTE). To enhance performance of model and to reduce the computation power needed, the feature selection was done using advanced optimization methods such as Genetic Algorithm (GA) and Particle Swarm Optimization (PSO). Multiple Metrics like Accuracy, Positive Predictive Value, Sensitivity, and F1Score were executed to assess the performance of various machine learning algorithms including Support Vector Machine (SVM), Random Forest (RF) and Artificial Neural Networks (ANN). The results proved that the ANN model had maximum accuracy of 98% after feature selection. This demonstrates the success of these techniques in boosting classification accuracy. Using micro-array, this study seeks out the opportunities of machine-learning enhancement in breast cancer examination.