I have a table with this features:

   'tourney_id', 'tourney_name', 'surface', 'draw_size', 'tourney_level',
   'tourney_date', 'match_num', 'winner_id', 'winner_seed', 'winner_entry',
   'winner_name', 'winner_hand', 'winner_ht', 'winner_ioc', 'winner_age',
   'winner_rank', 'winner_rank_points', 'loser_id', 'loser_seed',
   'loser_entry', 'loser_name', 'loser_hand', 'loser_ht', 'loser_ioc',
   'loser_age', 'loser_rank', 'loser_rank_points', 'score', 'best_of',
   'round', 'minutes', 'w_ace', 'w_df', 'w_svpt', 'w_1stIn', 'w_1stWon',
   'w_2ndWon', 'w_SvGms', 'w_bpSaved', 'w_bpFaced', 'l_ace', 'l_df',
   'l_svpt', 'l_1stIn', 'l_1stWon', 'l_2ndWon', 'l_SvGms', 'l_bpSaved',

You can see that they are duplicate features with the difference that one half is for winners e.i winner_age and the other half for losers e.i loser_age. I not sure which approach to take as is not a classification problem in the sense that I don't have clear binary labels (of course I could categorize some of them) or regression because I don't have a clear target (each feature is both winner and loser). I'm trying to do some EDA first but in order to do it I have to have a clear understanding on which approach to take to start with in the first place.

  • 1
    $\begingroup$ 1. what are you planning to do with this data? 2. when winner columns are populated then loser columns are null or they are also populated for that row? $\endgroup$ Jul 17 '20 at 19:00
  • $\begingroup$ Im trying to choose what type of algorithm apply whether is regression or classification. Both are populated. e.i winner_height = 1.80 and loser_height = 1.74 $\endgroup$
    – TOMAS
    Jul 17 '20 at 21:34
  • $\begingroup$ In addition to Brian answer, if you want to predict the height of person then it is a regression problem. However, I wonder if there is a correlation between height and winnder or loser! $\endgroup$ Jul 18 '20 at 8:25
  • $\begingroup$ Im not trying to predict height, Im trying to predict winners and losers based on the on the features from winners and losers. Because I have numeric features I think its gonna be regression. But I don't know which approach to take. Im new on ML. Do I have to subdivide the dataset with the winner features and loser features ? Maybe use seleKbest to select the best features. Also I want to make Exploratory Data Analysis first. $\endgroup$
    – TOMAS
    Jul 18 '20 at 14:04

If you are trying to predict the target of winner or loser, then it is a binary classification problem.

If you frame it as a binary classification, you'll have to reorganize the data as tidy data such that each row is an individual observation and each column is a feature.

  • $\begingroup$ So looking online I found that I could use the .melt() function in Pandas. Each column would be a feature and the columns would be winners and losers. Is this right ? $\endgroup$
    – TOMAS
    Jul 18 '20 at 21:00

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.