$4)$ Create a custom transformer in your pipeline that:

creates a new column in the data that sums the absences$\_$G$1,$ absences$\_$G$2,$ and absences$\_$G$3,$ data and then drops those three columns has a parameter that when equal to True, drops the G$1$ and G$2$ columns, and when False, leaves columns in the data $3)$ Fill in missing values or drop the rows or columns with missing values inside your pipeline. $5)$ Perform feature scaling on continuous numeric data in profile. $6)$ One$-$ hot encode normal or categorical data in a pipeline. $7)$ Create a column transformer to transform your numeric and categorical data. $8)$ Correctly transform your training data using the above data preparation steps and pipleines. I should have distinct sets of transformed training data: one containing the G$1/$ G$2$ and another without the G$1/$ G$2$ columns