【机器学习入门笔记15：BP神经网络逼近股票收盘价格2】20190218

2019-02-18  by 崔斐然  

import os
import tensorflow as tf
import numpy as np
import matplotlib.pyplot as plt

os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2'

# linespace 线性增长的（从1，到15，共15个数据)
date = np.linspace(1,15,15)

endPrice = np.array([2511.90,2538.26,2510.68,2591.66,2732.98,
                     2701.69,2701.29,2678.67,2726.50,2681.50,
                     2739.17,2715.07,2823.58,2864.90,2919.08])

beginPrice = np.array([2438.71,2500.88,2534.95,2512.52,2594.04,
                       2743.26,2697.47,2695.24,2678.23,2722.13,
                       2674.93,2744.13,2717.46,2832.73,2877.40])

plt.figure()
for i in range(0,15):
    # 1 柱状图
    dateOne = np.zeros([2])
    dateOne[0] = i;
    dateOne[1] = i;
    priceOne = np.zeros([2])
    priceOne[0] = beginPrice[i]
    priceOne[1] = endPrice[i]
    if endPrice[i]>beginPrice[i]:
        plt.plot(dateOne,priceOne,'r',lw=8)
    else:
        plt.plot(dateOne,priceOne,'g',lw=8)


#plt.show()
#思路：
# A(15x1)*w1(1x10)+b1(1*10) = B(15x10)
# B(15x10)*w2(10x1)+b2(15x1) = C(15x1)
# 1 A B C

# 归一化
dateNormal = np.zeros([15,1])
priceNormal = np.zeros([15,1])

for i in range(0,15):
    dateNormal[i,0] = i/14.0;
    priceNormal[i,0] = endPrice[i]/3000.0;

x = tf.placeholder(tf.float32,[None,1])
y = tf.placeholder(tf.float32,[None,1])

# B
# W1 b1 是可变的一个值
w1 = tf.Variable(tf.random_uniform([1,10],0,1))
b1 = tf.Variable(tf.zeros([1,10]))
wb1 = tf.matmul(x,w1)+b1
layer1 = tf.nn.relu(wb1) # 激励函数
# C
w2 = tf.Variable(tf.random_uniform([10,1],0,1))
b2 = tf.Variable(tf.zeros([15,1]))
wb2 = tf.matmul(layer1,w2)+b2
layer2 = tf.nn.relu(wb2)
# 真实值与 layer2 计算值得差异 计算均值。（开方再均值，其实就是标准差公式而已)）
loss = tf.reduce_mean(tf.square(y-layer2))
#train_step 表示每次调整的步长。使用的tf的梯度下降法GradientDescentOptimizer，被缩小参数为（loss）
train_step = tf.train.GradientDescentOptimizer(0.1).minimize(loss)
with tf.Session() as sess:
    sess.run(tf.global_variables_initializer())
    # 训练终止条件：方法1：for循环的次数  方法2、真实值与与测试的差值在允许范围内
    # 这里采用训练十万次的方法。
    for i in range(1,100000):
        sess.run(train_step,feed_dict={x:dateNormal,y:priceNormal})
    # w1w2 b1b2  A + wb -->layer2
    pred = sess.run(layer2,feed_dict={x:dateNormal})
    predPrice = np.zeros([15,1])
    # 打印预测结果
    for i in range(0,15):
        predPrice[i,0]=(pred*3000)[i,0]
    plt.plot(date,predPrice,'b',lw=1)
plt.show()