利用回归算法训练和预测

from sklearn.neural_network import MLPRegressor
from sklearn.externals import joblib
from sklearn import preprocessing
import numpy as np
import sys

inputn = int(sys.argv[1])#3
outputn = int(sys.argv[2])#1

inputcol = tuple(range(0,inputn,1))
print inputcol
X = np.loadtxt(open(sys.argv[3],"r"),delimiter="\t",usecols=inputcol)  
print X

outputcol = tuple(range(inputn,inputn+outputn,1))
print outputcol
y = np.loadtxt(open(sys.argv[3],"r"),delimiter="\t",usecols=outputcol)  
print y

sc = preprocessing.StandardScaler()
X_scaled = sc.fit_transform(X) 

clf = MLPRegressor(activation='logistic',solver='sgd',learning_rate='adaptive',)


clf.fit(X_scaled, y)   
print(clf.get_params(True))

filename = 'nbrbp.pkl'
joblib.dump(clf, filename, compress=9)

 

验证准确性

from sklearn.externals import joblib
from sklearn import preprocessing
import numpy as np
import sys

def maxidx(listfoo):
    #print(listfoo)
    vmax = max(listfoo)
    for i in range(0,len(listfoo)):
        if listfoo[i]==vmax:
            #print i
            return i
            
filename = 'nbrbp.pkl'
clf2 = joblib.load(filename)

inputn = int(sys.argv[1])
outputn = int(sys.argv[2])
inputcol = tuple(range(0,inputn,1))
outputcol = tuple(range(inputn,inputn+outputn,1))
X = np.loadtxt(open(sys.argv[3],"r"),delimiter="\t",usecols=inputcol)  
y = np.loadtxt(open(sys.argv[3],"r"),delimiter="\t",usecols=outputcol)  

sc = preprocessing.StandardScaler()
X_scaled = sc.fit_transform(X) 

hitmos1=0
hitmos2=0
hitmos3=0
maxidxv=0
hit=0
allcount=0
for i in range(0,X.shape[0]):    
    left = clf2.predict(X_scaled[i].reshape(1,-1)).tolist()
    right = y[i].tolist()
    tmpf = (float(left[0])-100)*(float(right)-100)
    if tmpf>0:
        flag = 1
    else:
        flag = 0    
    print str(left[0])+'\t'+str(right)+'\t'+str(flag)