Add manual indicators for the most prominent words in the feature hashing code by pre-constructing a partial hash table. Modify the function casl_fhash_matrix to account for this and re-run the elastic net using just the hash size of 16. How much do these custom defined terms effect the predictive power of the model on the testing set?
Already registered? Login
Not Account? Sign up
Enter your email address to reset your password
Back to Login? Click here