tensorflow 数据竞赛

最新推荐文章于 2024-02-02 23:56:18 发布

原创最新推荐文章于 2024-02-02 23:56:18 发布 · 601 阅读

0 ·

CC 4.0 BY-SA版权

Python 专栏收录该内容

82 篇文章

订阅专栏

本文介绍了一种基于TensorFlow的电力设备故障预测模型。该模型使用了标准化处理、特征工程和两层全连接神经网络进行训练。通过对训练数据集进行预处理并提取关键特征，模型能够准确预测电力设备的故障情况。

import pandas as pd
import sklearn.preprocessing as preprocessing

def make_data():
    col_names = ["ID","K1K2驱动信号","电子锁驱动信号","急停信号","门禁信号","THDV-M","THDI-M","label"]
    data = pd.read_csv("data_train.csv",names=col_names)
    # print(data.info())
    # print(data.describe())
    data = data.fillna(0)
    data['label2'] = data['label'].apply(lambda s: 1 - s)
    data["test1"]=data["THDI-M"]*data["THDV-M"]
    data["test2"]=data["急停信号"]*data["THDV-M"]
    data["test3"]=data["急停信号"]*data["THDV-M"]*data["THDI-M"]
    data["test4"]=data["THDI-M"]/data["THDV-M"]

    scaler=preprocessing.StandardScaler()
    lists=["K1K2驱动信号", "电子锁驱动信号", "急停信号", "门禁信号", "THDV-M", "THDI-M","test1","test2","test3"]
    for list in lists:
        data[list] = scaler.fit_transform(data[[list]])
    '''
    data["test1"] = scaler.fit_transform(data[["test1"]])
    data["test2"] = scaler.fit_transform(data[["test2"]])
    data["K1K2驱动信号"]=scaler.fit_transform(data[["K1K2驱动信号"]])
    data["电子锁驱动信号"] = scaler.fit_transform(data[["电子锁驱动信号"]])
    data["急停信号"] = scaler.fit_transform(data[["急停信号"]])
    data["门禁信号"] = scaler.fit_transform(data[["门禁信号"]])
    data["THDV-M"] = scaler.fit_transform(data[["THDV-M"]])
    data["THDI-M"] = scaler.fit_transform(data[["THDI-M"]])
    '''
    print(data)
    #print(data[["K1K2驱动信号","电子锁驱动信号","急停信号","门禁信号","THDV-M","THDI-M","label","label2"]])
    return data
def read_data():
    col_names = ["ID", "K1K2驱动信号", "电子锁驱动信号", "急停信号", "门禁信号", "THDV-M", "THDI-M",]
    data = pd.read_csv("data_test.csv", names=col_names)
   # print(data.info())
    data = data.fillna(0)
    data["test1"] = data["THDI-M"] * data["THDV-M"]
    data["test2"] = data["急停信号"] * data["THDV-M"]
    data["test3"] = data["急停信号"] * data["THDV-M"] * data["THDI-M"]
    data["test4"] = data["THDI-M"] / data["THDV-M"]
    scaler = preprocessing.StandardScaler()
    lists=["K1K2驱动信号", "电子锁驱动信号", "急停信号", "门禁信号", "THDV-M", "THDI-M","test1","test2","test3"]
    for list in lists:
        data[list] = scaler.fit_transform(data[[list]])

    '''
    data["test1"] = scaler.fit_transform(data[["test1"]])
    data["test2"] = scaler.fit_transform(data[["test2"]])
    data["K1K2驱动信号"] = scaler.fit_transform(data[["K1K2驱动信号"]])
    data["电子锁驱动信号"] = scaler.fit_transform(data[["电子锁驱动信号"]])
    data["急停信号"] = scaler.fit_transform(data[["急停信号"]])
    data["门禁信号"] = scaler.fit_transform(data[["门禁信号"]])
    data["THDV-M"] = scaler.fit_transform(data[["THDV-M"]])
    data["THDI-M"] = scaler.fit_transform(data[["THDI-M"]])
    '''
    return data[["K1K2驱动信号", "电子锁驱动信号", "急停信号", "门禁信号", "THDV-M", "THDI-M","test1","test2","test3","test4"]],data["ID"]
if __name__=="__main__":

    make_data()

import numpy as np
import pandas as pd
import tensorflow as tf
from sklearn.model_selection import train_test_split
import baidu_test
# read data from file
data = baidu_test.make_data()
# select features and labels for training
dataset_X = data[["K1K2驱动信号","电子锁驱动信号","急停信号","门禁信号","THDV-M","THDI-M","test1","test2","test3","test4"]].as_matrix()
dataset_Y = data[["label","label2"]].as_matrix()
print("--------------")
print(dataset_X)
# split training data and validation set data
X_train, X_val, y_train, y_val = train_test_split(dataset_X, dataset_Y,
                                                  test_size=0.2,
                                                  random_state=42)

# create symbolic variables
X = tf.placeholder(tf.float32, shape=[None, 10])
y = tf.placeholder(tf.float32, shape=[None, 2])


l1 = tf.layers.dense(X, 50, tf.nn.tanh, name="l1")
l2 = tf.layers.dense(l1, 20, tf.nn.tanh, name="l2")

out = tf.layers.dense(l2, 2, name="out")
y_pred = tf.nn.softmax(out, name="pred")
# weights and bias are the variables to be trained
loss = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(labels=y, logits=y_pred))
# 训练
train_step = tf.train.GradientDescentOptimizer(learning_rate=0.1).minimize(loss)
# 初始化变量
init = tf.global_variables_initializer()

# 结果存放在一个布尔型列表中
correct_prediction = tf.equal(tf.argmax(y, 1), tf.argmax(y_pred, 1))  # argmax返回一维张量中最大的值所在的位置
# 求准确率
accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))

# use session to run the calculation
with tf.Session() as sess:
    sess.run(init)
    for step in range(12000):
        sess.run(train_step, feed_dict={X: X_train, y: y_train})
        if step % 100 == 0:
            acc=sess.run(accuracy,feed_dict={X: X_val, y: y_val})
            print("acc is :"+str(acc))


    subdata,Id=baidu_test.read_data()
    #print(Id.as_matrix())
    prediction = np.argmax(sess.run(y_pred, feed_dict={X: subdata}),1)
    for i in range(len(prediction)):
        if prediction[i]==0:
            prediction[i]=1
        else:
            prediction[i]=0
    submission = pd.DataFrame({
        "ID": Id,
        "predictrion": prediction
    })
    submission.to_csv("baidu_sub2.csv",index=False)
    print("over")

import numpy as np
import pandas as pd
import tensorflow as tf
from sklearn.model_selection import train_test_split
import baidu_test
# read data from file
data = baidu_test.make_data()
# select features and labels for training
dataset_X = data[["K1K2驱动信号","电子锁驱动信号","急停信号","门禁信号","THDV-M","THDI-M","test1","test2","test3","test4"]].as_matrix()
dataset_Y = data[["label","label2"]].as_matrix()
print("--------------")
print(dataset_X)
# split training data and validation set data
X_train, X_val, y_train, y_val = train_test_split(dataset_X, dataset_Y,
                                                  test_size=0.2,
                                                  random_state=42)

# create symbolic variables
X = tf.placeholder(tf.float32, shape=[None, 10])
y = tf.placeholder(tf.float32, shape=[None, 2])


l1 = tf.layers.dense(X, 50, tf.nn.tanh, name="l1")
l2 = tf.layers.dense(l1, 20, tf.nn.tanh, name="l2")

out = tf.layers.dense(l2, 2, name="out")
y_pred = tf.nn.softmax(out, name="pred")
# weights and bias are the variables to be trained
loss = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(labels=y, logits=y_pred))
# 训练
train_step = tf.train.GradientDescentOptimizer(learning_rate=0.1).minimize(loss)
# 初始化变量
init = tf.global_variables_initializer()

# 结果存放在一个布尔型列表中
correct_prediction = tf.equal(tf.argmax(y, 1), tf.argmax(y_pred, 1))  # argmax返回一维张量中最大的值所在的位置
# 求准确率
accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))

# use session to run the calculation
with tf.Session() as sess:
    sess.run(init)
    step=0
    while True:
        step=step+1
        sess.run(train_step, feed_dict={X: X_train, y: y_train})
        if step % 100 == 0:
            acc=sess.run(accuracy,feed_dict={X: X_val, y: y_val})
            print("acc is :"+str(acc))
            if acc >0.89:
                break


    subdata,Id=baidu_test.read_data()
    #print(Id.as_matrix())
    prediction = np.argmax(sess.run(y_pred, feed_dict={X: subdata}),1)
    for i in range(len(prediction)):
        if prediction[i]==0:
            prediction[i]=1
        else:
            prediction[i]=0
    submission = pd.DataFrame({
        "ID": Id,
        "predictrion": prediction
    })
    submission.to_csv("baidu_sub2.csv",index=False)
    print("over")