sql-ann.rar_ANN_神经网络模型

# -*- coding: utf-8 -*- import pymssql import pandas as pd import numpy as np connect = pymssql.connect( host = '10.0.0.136\mssql', user = 'sa', password = 'PasswordGS15', database='test', charset='utf8') cursor = connect.cursor() sql='select * from dbo.Advertising' cursor.execute(sql) data = cursor.fetchall() df = pd.DataFrame(list(data),columns = ['No','TV','Radio','Newspaper','Sales']) train_df=df.head(20) #转换类型 TV=np.array(train_df['TV'].astype('float64')) Radio=np.array(train_df['Radio'].astype('float64')) Newspaper=np.array(train_df['Newspaper'].astype('float64')) Sales =np.array(train_df['Sales'].astype('float64')) #归一化过程 TV_=(TV-min(TV))/(max(TV)-min(TV)) Radio_=(Radio-min(Radio))/(max(Radio)-min(Radio)) Newspaper_=(Newspaper-min(Newspaper))/(max(Newspaper)-min(Newspaper)) Sales_=(Sales-min(Sales))/(max(Sales)-min(Sales)) train=np.vstack((TV_,Radio_,Newspaper_)) train_=np.transpose(train) Sales_1=np.vstack(Sales_) def nonlin(x,deriv=False): if(deriv==True): return x*(1-x) return 1/(1+np.exp(-x)) X = train_#10*3 y = Sales_1 #10*1 np.random.seed(1) syn0 = 2*np.random.random((3,5)) - 1 #初始化连接权 syn1 = 2*np.random.random((5,1)) - 1 for j in xrange(60000): l0 = X l1 = nonlin(np.dot(l0,syn0)) #10*3.*3*5=10*5 l2 = nonlin(np.dot(l1,syn1)) #10*1 l2_error = y - l2 if (j% 10000) == 0: print "Error:" + str(np.mean(np.abs(l2_error))) l2_delta = l2_error*nonlin(l2,deriv=True) #=10*1.* l1_error = l2_delta.dot(syn1.T)#= l1_delta = l1_error * nonlin(l1,deriv=True) syn1 += l1.T.dot(l2_delta) syn0 += l0.T.dot(l1_delta) print l2 print y.dtype #test test=[(218.4-min(TV))/(max(TV)-min(TV)),(27.7-min(Radio))/(max(Radio)-min(Radio)),(53.4-min(Newspaper))/(max(Newspaper)-min(Newspaper))] #1*3 l1_test=nonlin(np.dot(test,syn0)) #1*5 l2_test=nonlin(np.dot(l1_test,syn1)) #1*1 predict=l2_test*(max(Sales)-min(Sales))+min(Sales) print predict SQL="insert into predict values(s%)" cur.execute(SQL)