Suppose I have a Pandas dataframe, `df`

, which has the following structure:-

```
Column 1 Column 2 .... Column 100
Row 1 0.233 0.555 0
Row 2 0.231 0.514 2
..
Row 15000 0.232 0.455 3
```

`Column 100`

represents a particular class each row belongs to (which can be from `0-14`

). Each category/class has `1000`

rows associated to it. For each category (denoted by integers in `Column 100`

), I only want to select `200`

samples randomly, and create a new dataframe `df_new`

which will have a new dimension of `15x200 = 3000 rows`

. Any good way to reach this?