Text Classification: All Tips and Tricks from 5 Kaggle Competitions
In this article, I will discuss some great tips and tricks to improve the performance of your text classification model. These tricks are obtained from solutions of some of Kaggle’s top NLP competitions. Namely, I’ve gone through: Jigsaw Unintended Bias in Toxicity Classification – $65,000 Toxic Comment Classification Challenge – $35,000 Quora Insincere Questions Classification – $25,000 Google QU
kaggle: Porto Seguro's Safe Driver Prediction ã¾ã¨ã‚ - copypasteã®æ—¥è¨˜
ã¯ã˜ã‚ã« ã‚³ãƒ³ãƒšæ¦‚è¦ ãƒ‡ãƒ¼ã‚¿ã®ç¨®é¡žã¨ã‚¿ã‚¹ã‚¯ 評価方法 æ出方法 勉強ã«ãªã‚‹ Kernel 㨠Discussion Data Preparation & Exploration | Kaggle Stratified KFold+XGBoost+EDA Tutorial(0.281) | Kaggle Resampling strategies for imbalanced datasets | Kaggle Python target encoding for categorical features | Kaggle Dimensionality reduction (PCA, tSNE) | Kaggle Tune and compare XGB, LightGBM, RF with Hyperopt | Kaggle 2-level Stacker | Kaggle Entity
Kaggle Past Competitions
Kaggle Past Solutions Sortable and searchable compilation of solutions to past Kaggle competitions. If you are facing a data science problem, there is a good chance that you can find inspiration here! This page could be improved by adding more competitions and more solutions: pull requests are more than welcome. Warning: this is a work in progress, many competitions are missing solutions. If you f
Stacking - 計算編①- - Qiita
1. ã¯ã˜ã‚㫠データコンペ界隈主導ã§ã€æ•°%ã®ç²¾åº¦ã‚’高ã‚る手法ã¨ã—ã¦stacking・blendingãŒéžå¸¸ã«æµè¡Œã£ã¦ãŠã‚Šã¾ã™(言ã‚ãšã‚‚ãŒãªã§ã™ãŒ)。実務家ã¨ã—ã¦ç€ç›®ã—ãŸã„ã®ã¯ã€æ•°%ã®æ”¹å–„ã‚’ã™ã‚‹æ–¹æ³•è«–ã ã¨æ€ã„ã¾ã™ã€‚æ•°%ã§ãƒãƒ£ãƒªãƒãƒ£ãƒªåº¦åˆã¯å…¨ç„¶é•ã„ã¾ã™ã®ã§ã€æŠ€è¡“ã«ã‚ˆã‚Šåˆ©ç›Šè²¢çŒ®ã‚’ã™ã‚‹ã¨ã„ã†ç›®ç·šã§å¤§å¤‰é ¼ã‚‚ã—ã„手法ã§ã™ã€‚StackingãŒKaggle主導ã§èª•ç”Ÿã—ãŸç†ç”±ã‚‚ã€ä»¥ä¸‹ã‚’覗ãã¨ã†ãªãšã‘ã¾ã™ï¼› [http://goo.gl/T0OlsV より引用] 1ä½ã¨2ä½ã®å·®ãŒ2bp(âˆ10^-4)ã¨ã¯å…¨ãé©šãã§ã™ã€‚ãã“ã§æœ¬ã‚¨ãƒ³ãƒˆãƒªãƒ¼ã§ã¯ã€Stackingã®æ¦‚è¦ã‚’æ•´ç†ã—ãŸå¾Œã€sklearnを使ã£ã¦å®Ÿé¨“ã‚’è¡Œã„ã€Stackingã«ã‚ˆã‚Šå¾—られる示唆を考察ã—ã¦ã¿ãŸã„ã¨æ€ã„ã¾ã™ã€‚ 2. Stackingã¨ã¯ 実ã®ã¨ã“ã‚ã€ä½•ãŒæ£ã—ã„ã®ã‹ä¸æ˜Žçžãªã®ã§ã™ãŒã€[4]ã‚’å‚ç…§ã—解釈ã™ã‚‹ã¨è¦–点ã¨ã—ã¦ã¯ä»¥ä¸‹ãŒæŒ™ã’られるã¨è€ƒãˆã¦ã„ã¾
ã„ã‹ã«ã—ã¦kaggleを解ãã‹ | threecourse's memo
Walmart2015ã«ã¤ã„ã¦ã¯ã€CrowdFlowerã®Solutionã‚’å‚考ã«ã—ãªãŒã‚‰ã€ã„ã‚ã„ã‚ã¨è€ƒãˆãªãŒã‚‰ã‚„ã£ã¦ã¾ã—ãŸã€‚ çµå±€xgboostã¨neural netã®stackingãらã„ã—ã‹ã§ãã¾ã›ã‚“ã§ã—ãŸãŒã€‚。(38th/1047) ãã®ä¸ã§ã€kaggleã®å•é¡Œã«å¯¾ã—ã¦â€è§£â€ã‚’出ã™ã®ã«ã‚ãŸã£ã¦ã€ã©ã†ã„ã†ãƒã‚¤ãƒ³ãƒˆãŒã‚ã‚Šã€ã©ã®ã‚ˆã†ã«é€²ã‚ã¦ã„ãã¹ãã‹ã®â€æ§‹é€ â€ã‚’考ãˆã¦ã¿ã¾ã—ãŸã€‚ kaggle masterã ã‘ã©æ©Ÿæ¢°å¦ç¿’ã¯ç´ 人ãªã®ã§ã€çªã£è¾¼ã¿ã©ã“ã‚ã¯ã„ã£ã±ã„ã‚ã‚‹ã¨æ€ã„ã¾ã™ã€‚ã”指摘やã“ã‚“ãªè³‡æ–™ãŒã‚ã‚‹ãªã©ã€æ•™ãˆã¦ã„ãŸã ã‘ã‚Œã°åŠ©ã‹ã‚Šã¾ã™ã€‚ 1 – 特徴é‡ä½œæˆãƒ»ç®¡ç† 何ã ã‹ã‚“ã ã„ã£ã¦ã‚‚特徴é‡ä½œæˆã¯é‡è¦ã§ã€é‡è¦ãªç‰¹å¾´é‡ã‚’逃ã™ã¨å‹ã¡ç›®ã¯ãªã„ã§ã—ょã†ã€‚ 逆ã«é©åˆ‡ãªç‰¹å¾´é‡ãŒä½œã‚Œã‚Œã°ã€xgboostã ã‘ã§å‹ã¦ã‚‹ã“ã¨ã‚‚ã‚るよã†ã§ã™ã€‚(cf. Rossmann) 特徴é‡ã¯N * M（N:ID, M:è¦ç´ 数）ã®æ•°
1
リリースã€éšœå®³æƒ…å ±ãªã©ã®ã‚µãƒ¼ãƒ“スã®ãŠçŸ¥ã‚‰ã›
最新ã®äººæ°—エントリーã®é…ä¿¡
j次ã®ãƒ–ックマーク
kå‰ã®ãƒ–ックマーク
lã‚ã¨ã§èªã‚€
eコメント一覧を開ã
oページを開ã
{{#tags}}- {{label}}
{{/tags}}