Logistic回归 逻辑回归 练习——以2018建模校赛为数据源

import tensorflow as tf
import os
import numpy
import xlrd

XDATA = 0
YDATA = 0
one_hot_size = 0
M = 0


def readData():
    global XDATA, YDATA, one_hot_size, M
    workbook = xlrd.open_workbook(‘divdata.xlsx‘)
    booksheet = workbook.sheet_by_index(0)
    col = booksheet.ncols
    row = booksheet.nrows
    M = row
    tempcol = []
    for i in range(col - 1):
        tempcol = tempcol + booksheet.col_values(i)
    XDATA = numpy.array(tempcol).reshape(col - 1, row).T
    one_hot_size = int(max(booksheet.col_values(col - 1)))
    YDATA = numpy.zeros([row, one_hot_size])
    for i in range(row):
        YDATA[i, int(booksheet.cell_value(i, col - 1) - 1)] = 1


def getData(batch_size):
    ran = numpy.random.randint(0, M - 1, [batch_size])
    # print(ran)
    return XDATA[ran], YDATA[ran]


readData()
checkpoint_dir = ‘modelsave/‘
learning_rate = 0.0005
save_step = 100
total_step = 1000
batch_size = 1000
config = tf.ConfigProto()
config.gpu_options.allow_growth = True

x = tf.placeholder(tf.float32, [None, 10], name=‘x‘)
y_data = tf.placeholder(tf.float32, [None, 4], name=‘data‘)
# y = tf.Variable(tf.zeros(4,1), dtype=tf.float32,name=‘y‘)
# w = tf.Variable(tf.zeros([10, 4], dtype=tf.float32))
w = tf.Variable(numpy.zeros([10, 4]),dtype=tf.float32)
# b = tf.Variable(tf.zeros([1, 4], dtype=tf.float32))
b = tf.Variable(numpy.zeros([1,4]),dtype=tf.float32)
y = tf.nn.softmax(tf.matmul(x, w) + b)

loss = tf.reduce_mean(-tf.reduce_sum(y_data * tf.log(y), reduction_indices=1))  # 损失函数
optimizer = tf.train.GradientDescentOptimizer(learning_rate)  # 选择梯度下降的方法
train_op = optimizer.minimize(loss)  # 迭代的目标：最小化损失函数
sess = tf.InteractiveSession(config=config)  # 设置按需使用GPU
saver = tf.train.Saver()  # 用来存储训练结果

if not os.path.exists(checkpoint_dir):
    os.mkdir(checkpoint_dir)

#############################
# 读取并初始化:
ckpt = tf.train.get_checkpoint_state(checkpoint_dir)
if ckpt and ckpt.model_checkpoint_path:
    saver.restore(sess, ckpt.model_checkpoint_path)
else:
    sess.run(tf.global_variables_initializer())
##############################


print(sess.run(b))
for i in range(total_step):
    batch = getData(batch_size)
    # print(batch[0])
    # print(batch[1])
    sess.run(train_op, feed_dict={x: batch[0], y_data: batch[1]})
    if (i + 1) % save_step == 0:
        print(i + 1, sess.run(w), sess.run(b))
        saver.save(sess, checkpoint_dir + ‘model.ckpt‘, global_step=i + 1)  # 储存

writer = tf.summary.FileWriter(‘./my_graph‘, sess.graph)  # tensorboard使用

writer.close()
sess.close()

# 查看tensorboard的代码 在命令行输入：
# tensorboard --logdir=C:\Users\Rear82\PycharmProjects\MM_School_2018\my_graph

import tensorflow as tf
import os
import numpy


checkpoint_dir = ‘modelsave/‘
config = tf.ConfigProto()
config.gpu_options.allow_growth = True

x = tf.placeholder(tf.float32, [None, 10], name=‘x‘)
w = tf.Variable(numpy.zeros([10, 4]),dtype=tf.float32)
b = tf.Variable(numpy.zeros([1,4]),dtype=tf.float32)
y = tf.nn.softmax(tf.matmul(x, w) + b)

sess = tf.InteractiveSession(config=config)  # 设置按需使用GPU
saver = tf.train.Saver()  # 用来存储训练结果

if not os.path.exists(checkpoint_dir):
    os.mkdir(checkpoint_dir)

#############################
# 读取并初始化:
ckpt = tf.train.get_checkpoint_state(checkpoint_dir)
if ckpt and ckpt.model_checkpoint_path:
    saver.restore(sess, ckpt.model_checkpoint_path)
else:
    print("Can‘t find trained nn.")
##############################

jdata = [[167,86,72,71.5,76.5,90.5,119.4444444,120.2797203,112.4183007,95.02762431]]
print(jdata)
print(sess.run(y,feed_dict={x:jdata}))

sess.close()