python MKL SVM(多核学习)

python 实现MKL SVM(多核学习)+lasso降维处理

import math
import pandas as pd
from sklearn.linear_model import LassoCV
from sklearn.feature_selection import SelectFromModel
from sklearn import preprocessing
import numpy as np
import matplotlib.pyplot as plt
from sklearn import svm
from sklearn.model_selection import train_test_split        
import itertools
import datetime
from sklearn.model_selection import cross_val_score
"定义核函数"
"np.linalg.norm(求范数)"
"范数的定义 eg向量x=[2,4,8]T(转至)的范数为:||x||=根号(2*2+4*4+8*8)=9.165"
"math.exp(1)返回e的一次方"
def rbf(gamma=1.0):
 def rbf_fun(x1,x2):
  return math.exp((np.linalg.norm(x1-x2))*(-1.0*gamma))
 return rbf_fun
"x2.transpose()是对矩阵的转置"
def lin(offset=0):
 def lin_fun(x1,x2):
  return x1.dot(x2.transpose())+offset
 return lin_fun
"pow(x1.dot(x2.transpose())+offset,power)指的是对得到x1.dot(x2.transpose())+offset的power次方"
def poly(power=2,offset=0):
 def poly_fun(x1,x2):
  return pow(x1.dot(x2.transpose())+offset,power)
 return poly_fun
 def sig(alpha=1.0,offset=0):
 def sig_fun(x1,x2):
  return math.tanh(alpha*1.0*x1.dot(x2.transpose())+offset)
 return sig_fun
"根据输入X的大小构造核矩阵"
def kernel_matrix(x,kernel):
 mat=np.zeros((x.shape[0],x.shape[0]))
 for a in range(x.shape[0]):
  for b in range(x.shape[0]):
   mat[a][b]=kernel(x[a],x[b])
 return mat
".trace()得到矩阵的迹eg a=[[a11,a12],[a21,a22]] a的迹就是a11+a22的值"
"f_dot函数最后得到一个值"
def f_dot(kernel_mat1,kernel_mat2):
 return (kernel_mat1.dot(kernel_mat2.transpose())).trace()
def A(kernel_mat1,kernel_mat2):
 return (f_dot(kernel_mat1,kernel_mat2))/(math.sqrt(f_dot(kernel_mat1,kernel_mat1)*f_dot(kernel_mat2,kernel_mat2)))
'''
求betas
1.形成一个y行y列的矩阵yyT,由y*yT得到
2.通过kernel_matrix,得到对X数据进行核函数的映射后的矩阵，
和X的行数列数相同，设为data
3.通过f_dot函数,将data和y相乘，返回相乘得到的矩阵f_mat,再返回矩阵的迹，记为J
4.通过A函数将f_mat和yyT相乘得到的迹，再除以根号下(f_mat*f_matT)*(yyT*yyTT)得到的矩阵的迹
5.将不同核函数在第四步得到的值相加，得到deno值
6.得到使用不同核函数情况下的betas值，
通过A函数将f_mat和yyT相乘得到的迹，再除以根号下
(f_mat*f_matT)*(yyT*yyTT)得到的矩阵的迹，最后除以deno
就的到每个核函数的betas值了
betas值是每个核函数的比重
'''
def beta_finder(x,y,kernel_list):
    y=np.matrix(y)
    yyT=y.dot(y.transpose())
    deno=sum([A(kernel_matrix(x,kernel),yyT) for kernel in kernel_list])
    betas=[A(kernel_matrix(x,kernel),yyT)/deno for kernel in kernel_list]
    print (betas)
    return betas
"产生multi核"
'''
1.得到betas
2.生成矩阵XxY维的矩阵
3.得到不同核函数对X数据映射后的数据data，再乘以该核函数对应的beta值（比重），再
4.将上述得到的矩阵相加得到最融合的矩阵
'''
def multi_kernel_maker(x,y,kernel_list):
 betas=[float(b) for b in beta_finder(x,y,kernel_list)]
 #print " ",betas
 def multi_kernal(x1,x2):
  mat=np.zeros((x1.shape[0],x2.shape[0]))
  for a in range(x1.shape[0]):
   for b in range(x2.shape[0]):
    mat[a][b]=sum([betas[i]*kernel(x1[a],x2[b]) for i,kernel in enumerate(kernel_list)])
  return mat
 return multi_kernal
"制造多核"
#kernels = [lin(),lin(2),poly(),poly(3),poly(4),rbf(),rbf(1.5),sig(),sig(1.5)]
kernels = [lin(),poly(),rbf(),sig(),rbf(10)]
kernel_numbers=5
multi_kernels = [mult for mult in itertools.combinations(kernels, kernel_numbers)]#itertools.combinations迭代器eg。(combinations('ABC', 2))得到[('A', 'B'), ('A', 'C'), ('B', 'C')]      
"训练模型"
def mk_train(x_train,y_train,multi_kernels):
    y=[[t] for t in y_train[:]]
    #  y=[[t] for t in y_train[:,i]]
    for k_list in multi_kernels:
        mk_train_start_time=datetime.datetime.now()
        multi_kernel=multi_kernel_maker(x_train,y,k_list)
        print(k_list,'multi kernel maked! !')
        clf=svm.SVC(kernel=multi_kernel)
        results=cross_val_score(clf,x_train, y_train, scoring='accuracy',cv=10)
        print(results.mean())
        mk_train_end_time=datetime.datetime.now()
        print('mk_train_time:',(mk_train_end_time-mk_train_start_time).seconds,'seconds')
        "导入数据"       
file_path=r'C:\\Users\\DengBY\\Desktop\\liver_Lasso_features\\ALL_V.csv'
data = pd.read_csv(file_path)
a=pd.DataFrame(data)
X=a.values[:,1:597]
y=a.values[:,598]
min_max_scaler = preprocessing.MinMaxScaler()#范围0-1缩放标准化
X=min_max_scaler.fit_transform(X)
"基于Lasso的特征选择"
lsvc=LassoCV().fit(X, y)
model = SelectFromModel(lsvc, prefit=True)
X_lsvc = model.transform(X)
df_X_lsvc=pd.DataFrame(X_lsvc)
y=pd.DataFrame(y)
b=df_X_lsvc
objs=[b,y]
"features select 后的数据"
data=pd.concat(objs, axis=1, join='outer', join_axes=None, ignore_index=False,
               keys=None, levels=None, names=None, verify_integrity=False
'打乱数据，重新排序'               
data=data.sample(frac=1)
X=data.values[:,:14]
y=data.values[:,15]   
X_train,X_test,y_train,y_test=train_test_split(X,y,random_state=200)
print('model training starting')
mk_train(X_train,y_train,multi_kernels)
print('model training finishing')
#保存日志
#import sys
#f_handler=open('out.log', 'w')
#sys.stdout=f_handler