Assume that you are given a task to adapt and further improve basic machine learning algorithms to analyse some complex data. Please answer the following questions based on such a situation.

$($a$)$ If the specific goal of the task is to find outliers $($abnormal data$)$ in the online data, is this scenario more suitable for unsupervised learning or supervised learning? Please explain the reasons for choosing specific learning paradigm and describe the potential shortcomings of the other one.

$($b$)$ Please describe a specific unsupervised learning algorithm and describe the algorithm flow.

$($c$)$ We know that there is a phenomenon called overfitting in supervised learning. So$,$ is there any overfitting phenomenon in unsupervised learning? If yes, please explain how to judge and propose solutions to address the overfitting. If not, please explain why?