RNN预测股票数据---模型和参数

最新推荐文章于 2024-08-19 10:12:10 发布

Adm1rat1on

最新推荐文章于 2024-08-19 10:12:10 发布

阅读量2.3k

点赞数

分类专栏：机器学习

本文链接：https://blog.csdn.net/qq_35358021/article/details/86649018

版权

本文探讨了如何利用循环神经网络(RNN)进行股票市场预测，详细介绍了模型构建过程及关键参数设置，帮助读者理解RNN在金融数据分析中的应用。

摘要由CSDN通过智能技术生成

# 定义模型参数
class NetConfig():
	def __init__(self):
		self.rnn_unit = 10
		self.input_size = 7
		self.output_size = 1
		self.lr = 0.0006
		self.time_step = 20
		self.batch_size = 80
		self.weights = {
   
			"in":tf.Variable(tf.random_normal([self.input_size,self.rnn_unit])),
			"out":tf.Variable(tf.random_normal([self.rnn_unit,self.output_size]))
}
		self.biases = {
   
			"in":tf.Variable(tf.constant(0.1,shape=[self.rnn_unit,])),
			"out":tf.Variable(tf.constant(0.1,shape=[self.output_size,]))
}

# 定义训练模型
class MyModel():
	def __init__(self):
		self.sess = tf.Session()
		self.nc = NetConfig()
	def train_rnn(self,data,save_path,iter_num):
		weights = self.nc.weights
		biases = self.nc.biases
		input_size = self.nc.input_size
		rnn_unit = self.nc.rnn_unit
		output_size = self.nc.output_size
		time_step = self.nc.time_step
		batch_size = self.nc.batch_size
		lr = self.nc.lr
		sess = self.sess
		
		X = tf.placeholder(tf.float32,shape=