masafumi330 / kaggle_MMLM2022

kaggle "MMLM 2022" competition repo.
1 stars 0 forks source link

【Experiments】Run the notebook of "Adapting the 2nd Place Solution of the 2021 NCAAW Competition" #10

Open Guion-Rem24 opened 2 years ago

Guion-Rem24 commented 2 years ago

About

URL: https://www.kaggle.com/theoviel/using-last-year-s-2nd-place

Description

採用する特徴量

各試合の対戦チームごとに各特徴量のカラムを持ち,加えて SeedDiffWinRatioDiffGapAvgDiff538ratingDiffを算出.

モデル

選択肢

所感

会場のデータをdropしているのは意味があるんだろうか. 538ratingのデータがどこまで効いているかわからないが,このままでは効いてない感じ コード多分いくつか間違ってそう?

method score
not use strategy 0.5414
safe strategy 0.5657
risky strategy 0.6503