解决方案排名第 46 名及一些特征修正

我们在其中一个集成模型中使用了模型的 MLM 训练以及下述特征修正。PLS（伪标签）在某种程度上导致了 CV（交叉验证）过拟合，因此我们将其与非 PL 模型混合，以获得收益并中和过拟合部分。

我们最佳的最终解决方案是以下模型的混合：

• Deb v3 Large
• Deb large
• Deb v3 Large PL

次优的选定解决方案是以下模型的混合：

4 个模型 -> deb l + debv3 No PL + debv3 feature corrections No PL + deberta base PL

基于错误分析或预测中的缺失部分，我提出了以下特征修正。
这确实显著改善了一些折的 CV，但不幸的是 LB（排行榜）并没有看到那么大的提升。
如果有人想尝试将这些修正应用到他们的最佳解决方案中，请尝试一下并告诉我结果。

 print(features.iloc[132,2])
    features.iloc[132,2]='Vomiting-or-V'
    print(features.iloc[73,2])
    features.iloc[73,2]='Episodes-of-heart-racing-or-heart-poundings-or-palpitations'
    print(features.iloc[90,2])
    features.iloc[90,2]='Recent-upper-respiratory-infection-symptoms or-nasal-congestion'
    print(features.iloc[119,2])
    features.iloc[119,2]='Diminished-energy-or-feeling-drained-or-somnolence'
    print(features.iloc[76,2])
    features.iloc[76,2]='No-illicit-drug-use-or-No-toxic-habbits'
    print(features.iloc[6,2])
    features.iloc[6,2]='Adderall-use-or-Amphetamine'
    print(features.iloc[9,2])
    features.iloc[9,2]='heart-pounding-OR-heart-racing-or-palpitations'
    print(features.iloc[102,2])
    features.iloc[102,2]='LMP-2-months-ago-or-Last-menstrual-period-2-months-ago'
    print(features.iloc[[12,51,88],2])
    features.iloc[[12,51,88],2]='Male-or-M-or-Mr.'
    print(features.iloc[131,2]) 
    features.iloc[131,2]='Neck-pain-or-Pain-neck'
    print(features.iloc[135,2]) 
    features.iloc[135,2]='viral-symptoms-OR-rhinorrhea-OR-scratchy-throat(ST)'
    #new additions
    print(features.iloc[95,2]) 
    features.iloc[95,2]='No-shortness-of-breath-or-denies-dyspnea'
    print(features.iloc[63,2]) 
    features.iloc[63,2]='No-depressed-mood(good energy level)'
    print(features.iloc[2,2]) 
    features.iloc[2,2]='Chest-pressure-or-chest-pain'
    print(features.iloc[4,2]) 
    features.iloc[4,2]='Lightheaded-or-dizziness-or-presyncopal'
    #print(features.iloc[3,2]) 
    #features.iloc[3,2]='Intermittent-symptoms-or-random-episode'
    #print(features.iloc[74,2]) 
    #features.iloc[74,2]='Recent-visit-to-emergency-department(ED/ER)-with-negative-workup'
    print(features.iloc[97,2]) 
    features.iloc[97,2]='No-or-minor-relief-with-asthma-inhaler'
    print(features.iloc[82,2]) 
    features.iloc[82,2]='Feels-hot-OR-Feels-clammy-or-sweat'
    print(features.iloc[77,2]) 
    features.iloc[77,2]='Associated-nausea(N)'
    print(features.iloc[81,2]) 
    features.iloc[81,2]='Associated-throat-tightness-or-throat-narrowing'
    print(features.iloc[54,2]) 
    features.iloc[54,2]='No-blood-in-stool-or-BRBPR'
    print(features.iloc[129,2]) 
    features.iloc[129,2]='Myalgias-or-muscle-achiness'
    #print(features.iloc[50,2]) 
    #features.iloc[50,2]='2-to-3-beers-a-week-or-quit-or-few'
    print(features.iloc[142,2]) 
    features.iloc[142,2]='Subjective-fever-or-high-temperature'
    print(features.iloc[117,2]) 
    features.iloc[117,2]='Unsuccessful-napping-or-no-naps'
    print(features.iloc[[57,134],2]) 
    features.iloc[[57,134],2]='Nausea-or-N'
    #features.iloc[5,2]='No-hair-changes-