人工智能-项目实践-数据预处理-“中国工业大数据创新竞赛” 数据预处理与分类

from __future__ import print_function import tensorflow as tf import csv import random def add_layer(inputs, in_size, out_size, activation_function=None): # add one more layer and return the output of this layer Weights = tf.Variable(tf.random_normal([in_size, out_size]), dtype=tf.float32) biases = tf.Variable(tf.zeros([1, out_size]) + 0.1, dtype=tf.float32) Wx_plus_b = tf.matmul(inputs, Weights) + biases if activation_function is None: outputs = Wx_plus_b else: outputs = activation_function(Wx_plus_b) return outputs # Load train data with open(r'C:\Users\Yuhang Du\Desktop\ACT\python\15\Ldata_number_train_15.csv', 'r') as Ldata_train: train_reader = csv.reader(Ldata_train) data_train = [] for row in train_reader: del row[27] del row[0] row = [float(item) for item in row] data_train.append(row) random.shuffle(data_train) # Load test data with open(r'C:\Users\Yuhang Du\Desktop\ACT\python\15\Ldata_number_test_15.csv', 'r') as Ldata_test: # 时间按序号排序的带标签测试数据集 test_reader = csv.reader(Ldata_test) x_test = [] y_test = [] for row in test_reader: del row[27] del row[0] row = [float(item) for item in row] x_test.append(row) random.shuffle(x_test) len_ = len(x_test) for i in range(len_): if float(x_test[i][26]) == 1: y_test.append([0, 1]) else: y_test.append([1, 0]) del x_test[i][26] # define placeholder for inputs to network x = tf.placeholder(tf.float32, [None, 26]) y_ = tf.placeholder(tf.float32, [None, 2]) # add hidden layer l1 = add_layer(x, 26, 100, activation_function=tf.nn.relu) l2 = add_layer(l1, 100, 100, activation_function=tf.nn.relu) l3 = add_layer(l2, 100, 40, activation_function=tf.nn.relu) # add output layer logits = add_layer(l3, 40, 2, activation_function=None) y = tf.nn.softmax(logits) # the error between prediction and real data cross_entropy = tf.nn.softmax_cross_entropy_with_logits(labels=y_, logits=logits) loss = tf.reduce_mean(cross_entropy) train_step = tf.train.GradientDescentOptimizer(0.01).minimize(loss) correct_prediction = tf.equal(tf.argmax(y, 1), tf.argmax(y_, 1)) accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32)) init = tf.global_variables_initializer() sess = tf.Session() sess.run(init) times = len_ // 256 counter = 0 for count in range(10000): # 训练集生成 x_batch_train = [] y_batch_train = [] for i in range(256): row = [data_train[i][j] for j in range(26)] x_batch_train.append(row) if data_train[i][26] == 1: y_batch_train.append([0, 1]) # 异常是[0,1] else: y_batch_train.append([1, 0]) # 正常是[1,0] train_feed_dict = {x: x_batch_train, y_: y_batch_train} sess.run(train_step, feed_dict=train_feed_dict) counter += 1 # if counter == times: random.shuffle(data_train) # counter = 0 if count % 10 == 0: print(count/10) print(sess.run(loss, feed_dict=train_feed_dict)) print(sess.run(accuracy, feed_dict=train_feed_dict)) print("Train step finished.\nThe loss and accuracy of test set are as followed:") # 输出预测信息 test_feed_dict = {x: x_test, y_: y_test} prediction = sess.run(y, feed_dict=test_feed_dict) with open(r'C:\Users\Yuhang Du\Desktop\ACT\python\15\Ldata_number_test_15.csv', 'r') \ as Ldata_test: with open(r'C:\Users\Yuhang Du\Desktop\ACT\python\15\Ldata_number_test_predict_15.csv', 'w', newline='') \ as Ldata_test_predict: # 带神经网络预测标签的测试数据集 reader = csv.reader(Ldata_test) writer = csv.writer(Ldata_test_predict) counter = 0 for row in reader: flag = 0 if prediction[counter][0] > prediction[counter][1]: pass else: flag = 1 row.append(flag) writer.writerow(row) counter = counter + 1 print(sess.run(loss, feed_dict=test_feed_dict)) print(sess.run(accuracy, feed_dict=test_feed_dict)) # abs # matplot # 错误信息