You are given a dataset for classifying songs into classical and non$-$classical. The size of the data is given in Table $1.$ Out of $5,000$ samples, there are $100$ samples of classical songs $($class label $0),$ and the remaining are nonclassical $($class label $1).$ You are asked to sample training data for building a classification model. What sampling technique will you use and justify how your approach will help reduce the training error?

$\backslash$table$[[$Songid$,$F$1,$F$2,$dots...,F$10,\backslash$table$[[$Class$],[$Label$]]],[1,,,,0,],[2,,,,,1],[.,,,,,],[5000,,,,,1]]$

Table $1$