The malware detection project consist of two parts. The first part is about binary-classification problem where the Target class has two labels 'benign' and 'gafgyt_combo'. In this part I compared the performance of two models: Logistic Regression and Decision Tree Classifier. The latter was the winner. I thought maybe the LR model performed poorly due to the existense of redundant features, so I removed them and re-evaluated the models' performance, but nothing has changed, wo I ended up with an open question is, should we always remove the highly correlated features?
The second part is a mutli-class classification problem, where I will be having 11 labels in my target class..to be continued...