神经网络dnn 多分类模型

import tensorflow.compat.v1 as tf
# from tensorflow.examples.tutorials.mnist import input_data
import os
import pandas as pd
import numpy as np
from tensorflow.python.keras.utils import to_categorical

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


# 数据准备
df = pd.read_csv('./data/train_date_new.csv',sep=',',index_col=None,header=0)

# print(df)

# print(df.groupby(by="diabete").count())

X = df.iloc[:,1:].values.astype(np.float32)
Y = to_categorical(df.iloc[:,0].values).astype(np.float32)


print(X.shape,Y.shape)

train_split = int(df.shape[0]*0.8)
x_train,y_train,x_test,y_test = X[:train_split,:],Y[:train_split,:],X[train_split:,:],Y[train_split:,:]

ind,col = x_train.shape
y_ind,y_col = y_train.shape

# 全连接神经网络
def dense(x, w, b, keeppord):
    linear = tf.matmul(x, w) + b
    # activation = tf.nn.relu(linear)
    activation = tf.nn.sigmoid(linear)
    # activation = tf.nn.tanh(linear)
    # activation = tf.nn.softmax(linear)
    y = tf.nn.dropout(activation,keeppord)
    return y


def DNNModel(image, w, b, keeppord):
    global dense1
    for i in range(len(w)-1):
        if i==0:
            dense1 = dense(image, w[i], b[i],keeppord)
        else:
            dense1 = dense(dense1, w[i], b[i],keeppord)

    output = tf.matmul(dense1, w[-1]) + b[-1]
    return output

# 生成网络的权重
def gen_weights(unit_list):
    w = []
    b = []
    # 遍历层数
    for i in range(len(unit_list)-1):
        sub_w = tf.Variable(tf.random_normal(shape=[unit_list[i], unit_list[i+1]]))
        sub_b = tf.Variable(tf.random_normal(shape=[1,unit_list[i+1]]))
        w.append(sub_w)
        b.append(sub_b)
    return w, b


x = tf.placeholder(tf.float32, [None, col])
y = tf.placeholder(tf.float32, [None, y_col])
keepprob = tf.placeholder(tf.float32)

global_step = tf.Variable(0)

# unit_list = [784, 512, 256, 10]
unit_list = [col, 512,256, y_col] #  0.7543333

# unit_list = [col,1024,512,y_col]
duropt = 0.75

w, b = gen_weights(unit_list)
y_pre = DNNModel(x, w, b, keepprob)
loss = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits_v2(logits=y_pre, labels=y))
tf.summary.scalar("loss", loss)                 # 收集标量

opt = tf.train.AdamOptimizer(0.01).minimize(loss, global_step=global_step)

predict = tf.equal(tf.argmax(y_pre, axis=1), tf.argmax(y, axis=1))       # 返回每行或者每列最大值的索引，判断是否相等
acc = tf.reduce_mean(tf.cast(predict, tf.float32))

tf.summary.scalar("acc", acc)                   # 收集标量
merged = tf.summary.merge_all()                 # 和并变量
saver = tf.train.Saver()                        # 保存和加载模型
init = tf.global_variables_initializer()        # 初始化全局变量



bach = 4
bach_0=bach-1
min_bach = int(ind/4)
print(bach_0,min_bach)


with tf.Session() as sess:
    sess.run(init)
    writer = tf.summary.FileWriter("./log/tensorboard", tf.get_default_graph())      # tensorboard 事件文件
    for i in range(10000):
        for j in range(bach):
            if j <= bach_0:
                x_train_bach, y_train_bach = x_train[(j * min_bach):(j + 1) * min_bach, :],\
                                             y_train[(j * min_bach):(j + 1) * min_bach,:]
            else:
                x_train_bach, y_train_bach = x_train[(j + 1) * min_bach:, :], y_train[(j + 1) * min_bach:, :]

            summary, _ = sess.run([merged, opt], feed_dict={x:x_train_bach, y:y_train_bach, keepprob: duropt})
            writer.add_summary(summary, i)              # 将每次迭代后的变量写入事件文件

        # 评估模型在验证集上的识别率
        if (i+1) % 1000 == 0:
            feeddict = {x: x_test, y: y_test, keepprob: 1.}      # 验证集
            valloss, accuracy = sess.run([loss, acc], feed_dict=feeddict)
            print(i, 'th batch val loss:', valloss, ', accuracy:', accuracy)

    saver.save(sess, './model/tfdnn.ckpt')        # 保存模型
    print('测试集准确度:', sess.run(acc, feed_dict={x:x_test, y:y_test, keepprob:1.}))

writer.close()