cheetahbright / tsa-decision-trees

Decision tree implementation on a data set from the Transporation Security Administration.
0 stars 0 forks source link

Linear Regression to Reduce Variables #4

Open malctaylor15 opened 6 years ago

malctaylor15 commented 6 years ago

Use different versions of elastic search and linear regression to reduce the number of variables

After one hot encoding, there are too many variables to choose from.

Find a way to drastically reduce the number of variables and increase the performance of a linear regression.

Some potential ideas:

  1. Use Lasso and Ridge regression to reduce the number of variables to something more managable and that will yield to better results.

  2. Use a stepwise method .. Forward or backwards regression

  3. Look into different linear regression types ... least squares, hinge etc.

More notes to come....