# Feature Selection Phase

I am trying to predict the overall age of an opportunity (creation date - closing date) this is my response variable

lets say an opportunity passes through 3 stages to close

For example: Opp x stayed in

• stage 1 : 30 days
• stage 2 : 10 days
• stage 3: 20 days

At stage 3 I might close it same date or wait some time

so if I waited some time to close, it will be createdon: 22/11/2018 & closedon:9/2/2019

There is opp y , where i close it in same date of stage 3, so createdon:22/11/2018 and closedate: 21/1/2019

Summary

+---------+--------+--------+--------+--------+
| OppName | oppAge | stage1 | stage2 | stage3 |
+---------+--------+--------+--------+--------+
| x       |     79 |     30 |     10 |     20 |
| y       |     60 |     30 |     10 |     20 |
+---------+--------+--------+--------+--------+


my question is :

1. Can I include stage1,2,3 as my independent variables to create a regression model?
2. They seem to nearly make the model ideal, so is it better to include maybe only stage 1? without 2 &3

• I transformed stages to a categorical nature, for example <30 days to 1 month 30-60 day transformed it to 1-2months and so on

 +-----------+
|  stage1   |
+-----------+
| <1month   |
| 1~2months |
| 6~7months |
+-----------+

• Then I did one-hot encode to the stages like stage 1

• Then I stopped ,wasn't sure whether to include everything or what?

• Noticed your edits! Instead of renaming them to something categorical like <1month, rename it to something like 0 for your <1 month category, 1 for 1~2months category, and so on. PCA not really required now. – Random Nerd Nov 22 '18 at 10:19