树分类器的极高精度度量

我正在做一个项目,将总统辩论推文分为中立、正面和负面。(不是当前的辩论数据集)。我正在训练和使用 Decision trees , Decision tree ensemble 和 AdaBoost

我拥有的数据是以 bag-of-words df_obama 是一个数据框架,包含所有关于奥巴马的推文。

df_Obama = pd.DataFrame.from_csv("../data/Obama_BagOfWords.csv")
df_Obama = df_Obama.reindex(np.random.permutation(df_Obama.index)).reset_index()
dataFeatures = df_Obama[allAttribs_Obama]
targetVar = list(df_Obama['Class'])

splitRatio = 0.9
splitPoint = int(splitRatio*len(dataFeatures))
dataFeatures_train = dataFeatures[:splitPoint]
dataFeatures_test = dataFeatures[splitPoint:]

targetVar_train = targetVar[:splitPoint]
targetVar_test = targetVar[splitPoint:]

clfObj = tree.DecisionTreeClassifier()
clfObj.fit(dataFeatures_train,targetVar_train)
preds = list(clfObj.predict(dataFeatures_test))
accScore = accuracy_score(targetVar_test,preds)
labels = [1,-1,0]

precision = precision_score(targetVar_test,preds,average=None,labels=labels)
recall = recall_score(targetVar_test,preds,average=None,labels=labels)
f1Score = f1_score(targetVar_test,preds,average=None,labels=labels)

print("Overall Acurracy",accScore)
print("precision",precision)
print("recall",recall)
print("f1Score",f1Score)

Overall Acurracy 1.0
precision [ 1.  1.  1.]
recall [ 1.  1.  1.]
f1Score [ 1.  1.  1.]

我就是不明白为什么会这样?指标如此之高有什么原因吗? 我还尝试了不同的列车测试分流比,结果似乎没有什么不同。

注: 以下是数据信息:

df_Obama.info()
<class 'pandas.core.frame.DataFrame'>
RangeIndex: 5465 entries, 0 to 5464
Columns: 13078 entries, level_0 to zzzzzzzzzz
dtypes: int64(13078)
memory usage: 545.3 MB

df_Obama.head(3)
0023Washington  08hayabusa  09Its   .... 09what 1000000th   
0               1           0            1       0
1               0           0            0       0
0               0           0            0       0