摘要
在日常生活中,我们常常会碰到各种各样的时间序列预测问题,如图书馆人流的预测、商品销量的预测、股价的预测,等等。TensorFlow Time Series(以下简称TFTS)是TensorFlow 1.3版本中新引入的模块。TFTS专门设计一套针对时间序列预测问题的API,利用其提供的LSTM模型,可以实现在TensorFlow中快速搭建高性能时间序列预测系统。
In daily life, we often encounter a variety of time series prediction problems, such as the prediction of library flow, the prediction of product sales, the prediction of stock prices and so on. The TensorFlow Time Series(hereafter referred to as TFTS)is a newly introduced module in the TensorFlow 1.3 release. TFTS has designed a set of APIs for time series prediction problems. Using its LSTM model, it can quickly build a high-performance time series prediction system in TensorFlow.
引文
[1]递归神经网络(Recurrent Neural Networks,RNN).https://www.cnblogs.com/ooon/p/5594428.html.
[2]王海静.基于HMM遗传算法的预测方法研究.黑龙江:哈尔滨工程大学,2014.
[3]Abadi M,Agarwal A,Barham P,et al.TensorFlow:Large-Scale Machine Learning on Heterogeneous Distributed Systems[J].arXiv:1603.04467[cs.DC].
[4]TensorFlow中文系列教程.http://www.tensorflownews.com/2017/08/30/tensorflow-time-series/.
[5]Abadi,M.,Barham,P.,Chen,J.,Chen,Z.,Davis,A.,Dean,J.,et al.(2016).TensorFlow:a System for Large-Scale Machine Learning.In OSDI(Vol.16:265-283).
[6]使用Estimator构建卷积神经网络.https://tensorflow.google.cn/tutorials/estimators/cnn.