Given a real-world binary classification problem (e.g. disease diagnosis: diseased or healthy) to be solved using a machine learning solution, state the key factors you would consider and methods you would apply in terms of data pre-processing and representa- tion (assuming data has been collected), feature engineering , model selection and evaluation process .

Given a dataset that contains 100 instances, in which each instance has 50 attributes (features) and a binary outcome event, describe what the potential problem is when you are asked to build a classification machine learning model based on this dataset. Describe how this problem can be addressed.

1. Describe the difference between intrinsic parameter and hyper-parameter in machine learn- ing. A specific example could be used to explain.

2. For a classification problem, we can use true positive (TP), false positive (FP), true negative (TN) and false negative (FN) to calculate a set of metrics (e.g. prediction accuracy, sen- sitivity, specificity etc.) for method evaluation. State the potential problem of only using prediction accuracy as the evaluation metric and how this problem can be addressed.