隨機森林數實現分類
數據預處理看代碼地址有操作指南
代碼地址? https://github.com/mastercaojie/Machine-Learning
# -*- coding: utf-8 -*-
import numpy as np? # numpy庫
from sklearn import cross_validation
from sklearn.model_selection import cross_val_score? # 交叉檢驗
import pandas as pd? # 導入pandas
from sklearn.ensemble import RandomForestClassifier
#獲取數據
data = pd.read_csv('corrected_new.csv',header=None,delimiter=",")
dataset = np.array(data)
print("數據集shape: ",dataset.shape)
print (70 * '-')? # 打印分隔線
X = dataset[:,0:35]
Y = dataset[:,35]
Label = np.array(['0','1','2','3','4'])
print("輸出類別:",Label)
print (70 * '-')? # 打印分隔線
# 劃分訓練集和測試集
X_train, X_test, Y_train, Y_test = cross_validation.train_test_split(X, Y, test_size=0.1)
print("訓練集個數：",X_train.shape[0])
print("測試集個數：",X_test.shape[0])
print (70 * '-')? # 打印分隔線
#用隨機森林
rf = RandomForestClassifier(bootstrap=True, class_weight=None, criterion='gini',
? ? ? ? ? ? max_depth=None, max_features='auto', max_leaf_nodes=None,
? ? ? ? ? ? min_samples_leaf=1, min_samples_split=2,
? ? ? ? ? ? min_weight_fraction_leaf=0.0, n_estimators=10, n_jobs=1,
? ? ? ? ? ? oob_score=False, random_state=None, verbose=0,
? ? ? ? ? ? warm_start=False)
rf.fit(X_train,Y_train)
probability = pd.DataFrame(np.array(rf.predict_proba(X_test)))
result = pd.DataFrame(np.array((rf.predict(X_test))))
score = cross_val_score(rf, X_train,Y_train)
print("訓練集精度得分:",score.mean())
score = cross_val_score(rf, X_test,Y_test)
print("測試集精度得分:",score.mean())
showlist = pd.concat([probability,result],axis=1)
print (70 * '-')? # 打印分隔線
print("概率+類別：")
showlist.columns = ['類別0','類別1','類別2','類別3','類別4',"結果"]
prob0 = showlist[showlist['結果'].isin([0])].shape[0]
prob1 = showlist[showlist['結果'].isin([1])].shape[0]
prob2 = showlist[showlist['結果'].isin([2])].shape[0]
prob3 = showlist[showlist['結果'].isin([3])].shape[0]
prob4 = showlist[showlist['結果'].isin([4])].shape[0]
resultlist =[prob0,prob1,prob2,prob3,prob4]
resultlist = pd.DataFrame(resultlist)
print(showlist)
print (70 * '-')? # 打印分隔線
print("統計每個類別的數量：")
resultlist.columns = ['統計']
print(resultlist)
運行結果
