登录
首页精彩阅读Python实现的随机森林算法与简单总结
Python实现的随机森林算法与简单总结
2018-02-15
收藏


Python实现的随机森林算法与简单总结

本文实例讲述了Python实现的随机森林算法。分享给大家供大家参考,具体如下:

随机森林数据挖掘中非常常用的分类预测算法,以分类或回归的决策树为基分类器。算法的一些基本要点:

*对大小为m的数据集进行样本量同样为m的有放回抽样;
*对K个特征进行随机抽样,形成特征的子集,样本量的确定方法可以有平方根、自然对数等;
*每棵树完全生成,不进行剪枝;
*每个样本的预测结果由每棵树的预测投票生成(回归的时候,即各棵树的叶节点的平均)

出于个人研究和测试的目的,基于经典的Kaggle 101泰坦尼克号乘客的数据集,建立模型并进行评估。

泰坦尼克号的沉没,是历史上非常著名的海难。突然感到,自己面对的不再是冷冰冰的数据,而是用数据挖掘的方法,去研究具体的历史问题,也是饶有兴趣。言归正传,模型的主要的目标,是希望根据每个乘客的一系列特征,如性别、年龄、舱位、上船地点等,对其是否能生还进行预测,是非常典型的二分类预测问题。数据集的字段名及实例如下:



值得说明的是,SibSp是指sister brother spouse,即某个乘客随行的兄弟姐妹、丈夫、妻子的人数,Parch指parents,children

下面给出整个数据处理及建模过程,基于ubuntu+python 3.4( anaconda科学计算环境已经集成一系列常用包,pandas numpy sklearn等,这里强烈推荐)

懒得切换输入法,写的时候主要的注释都是英文,中文的注释是后来补充的:-)

# -*- coding: utf-8 -*-
"""
@author: kim
"""
frommodelimport*#载入基分类器的代码
#ETL:same procedure to training set and test set
training=pd.read_csv('train.csv',index_col=0)
test=pd.read_csv('test.csv',index_col=0)
SexCode=pd.DataFrame([1,0],index=['female','male'],columns=['Sexcode']) #将性别转化为01
training=training.join(SexCode,how='left',on=training.Sex)
training=training.drop(['Name','Ticket','Embarked','Cabin','Sex'],axis=1)#删去几个不参与建模的变量,包括姓名、船票号,船舱号
test=test.join(SexCode,how='left',on=test.Sex)
test=test.drop(['Name','Ticket','Embarked','Cabin','Sex'],axis=1)
print('ETL IS DONE!')
#MODEL FITTING
#===============PARAMETER AJUSTMENT============
min_leaf=1
min_dec_gini=0.0001
n_trees=5
n_fea=int(math.sqrt(len(training.columns)-1))
#==============================================
'''''
BEST SCORE:0.83
min_leaf=30
min_dec_gini=0.001
n_trees=20
'''
#ESSEMBLE BY RANDOM FOREST
FOREST={}
tmp=list(training.columns)
tmp.pop(tmp.index('Survived'))
feaList=pd.Series(tmp)
fortinrange(n_trees):
#  fea=[]
  feasample=feaList.sample(n=n_fea,replace=False)#select feature
  fea=feasample.tolist()
  fea.append('Survived')
#    feaNew=fea.append(target)
  subset=training.sample(n=len(training),replace=True)#generate the dataset with replacement
  subset=subset[fea]
#  print(str(t)+' Classifier built on feature:')
#  print(list(fea))
  FOREST[t]=tree_grow(subset,'Survived',min_leaf,min_dec_gini)#save the tree
#MODEL PREDICTION
#======================
currentdata=training
output='submission_rf_20151116_30_0.001_20'
#======================
prediction={}
forrincurrentdata.index:#a row
  prediction_vote={1:0,0:0}
  row=currentdata.get(currentdata.index==r)
  forninrange(n_trees):
    tree_dict=FOREST[n]#a tree
    p=model_prediction(tree_dict,row)
    prediction_vote[p]+=1
  vote=pd.Series(prediction_vote)
  prediction[r]=list(vote.order(ascending=False).index)[0]#the vote result
result=pd.Series(prediction,name='Survived_p')
#del prediction_vote
#del prediction
#result.to_csv(output)
t=training.join(result,how='left')
accuracy=round(len(t[t['Survived']==t['Survived_p']])/len(t),5)
print(accuracy)

上述是随机森林的代码,如上所述,随机森林是一系列决策树的组合,决策树每次分裂,用Gini系数衡量当前节点的“不纯净度”,如果按照某个特征的某个分裂点对数据集划分后,能够让数据集的Gini下降最多(显著地减少了数据集输出变量的不纯度),则选为当前最佳的分割特征及分割点。代码如下:



# -*- coding: utf-8 -*-
"""
@author: kim
"""
importpandas as pd
importnumpy as np
#import sklearn as sk
importmath
deftree_grow(dataframe,target,min_leaf,min_dec_gini):
  tree={}#renew a tree
  is_not_leaf=(len(dataframe)>min_leaf)
  ifis_not_leaf:
    fea,sp,gd=best_split_col(dataframe,target)
    ifgd>min_dec_gini:
      tree['fea']=fea
      tree['val']=sp
#      dataframe.drop(fea,axis=1) #1116 modified
      l,r=dataSplit(dataframe,fea,sp)
      l.drop(fea,axis=1)
      r.drop(fea,axis=1)
      tree['left']=tree_grow(l,target,min_leaf,min_dec_gini)
      tree['right']=tree_grow(r,target,min_leaf,min_dec_gini)
    else:#return a leaf
      returnleaf(dataframe[target])
  else:
    returnleaf(dataframe[target])
  returntree
defleaf(class_lable):
  tmp={}
  foriinclass_lable:
    ifiintmp:
      tmp[i]+=1
    else:
      tmp[i]=1
  s=pd.Series(tmp)
  s.sort(ascending=False)
  returns.index[0]
defgini_cal(class_lable):
  p_1=sum(class_lable)/len(class_lable)
  p_0=1-p_1
  gini=1-(pow(p_0,2)+pow(p_1,2))
  returngini
defdataSplit(dataframe,split_fea,split_val):
  left_node=dataframe[dataframe[split_fea]<=split_val]
  right_node=dataframe[dataframe[split_fea]>split_val]
  returnleft_node,right_node
defbest_split_col(dataframe,target_name):
  best_fea=''#modified 1116
  best_split_point=0
  col_list=list(dataframe.columns)
  col_list.remove(target_name)
  gini_0=gini_cal(dataframe[target_name])
  n=len(dataframe)
  gini_dec=-99999999
  forcolincol_list:
    node=dataframe[[col,target_name]]
    unique=node.groupby(col).count().index
    forsplit_pointinunique:#unique value
      left_node,right_node=dataSplit(node,col,split_point)
      iflen(left_node)>0andlen(right_node)>0:
        gini_col=gini_cal(left_node[target_name])*(len(left_node)/n)+gini_cal(right_node[target_name])*(len(right_node)/n)
        if(gini_0-gini_col)>gini_dec:
          gini_dec=gini_0-gini_col#decrease of impurity
          best_fea=col
          best_split_point=split_point
    #print(col,split_point,gini_0-gini_col)
  returnbest_fea,best_split_point,gini_dec
defmodel_prediction(model,row):#row is a df
  fea=model['fea']
  val=model['val']
  left=model['left']
  right=model['right']
  ifrow[fea].tolist()[0]<=val:#get the value
    branch=left
  else:
    branch=right
  if('dict'instr(type(branch) )):
    prediction=model_prediction(branch,row)
  else:
    prediction=branch
  returnprediction

实际上,上面的代码还有很大的效率提升的空间,数据集不是很大的情况下,如果选择一个较大的输入参数,例如生成100棵树,就会显著地变慢;同时,将预测结果提交至kaggle进行评测,发现在测试集上的正确率不是很高,比使用sklearn里面相应的包进行预测的正确率(0.77512)要稍低一点 :-(  如果要提升准确率,两个大方向: 构造新的特征;调整现有模型的参数。

这里是抛砖引玉,欢迎大家对我的建模思路和算法的实现方法提出修改意见。



数据分析咨询请扫描二维码

客服在线
立即咨询