在股票市场大力提倡价值投资的背景下，准确预测公司未来收入和经营现金流，是理性投资者最重要的事情。买入盈利超预期的公司，避开盈利能力差的公司，才是投资的长久之道，但是在A股市场，盈利容易被操纵，存在不确定性，但是经营活动现金流不便于捏造，而且该指标反映了企业的生产经营实际状况，所以本文选择使用机器学习算法预测此指标，进而预测企业的未来发展趋势。经营现金流定义为：企业在生产经营活动中，因提供各种服务或者销售产品而取得的各项收入，它关系到企业的生存和发展，对企业有重要的意义。本文先介绍了机器学习的基本思想、主要方法、最新进展以及应用。然后介绍机器学习预测企业经营现金流的框架，本文尝试了多种方法，包括LR、XGBOOSTOOST、DNN等算法，以及bagging、stacking等组合方法，在实验的过程中，使用进行了基于梯度下降算法、滑窗算法的效果优化，基于max-min、标准正态等方式的特征预处理等方法，最终实现了一个较好的预测效果。

Under the background that the stock market strongly advocates value investment, it is the most important thing for rational investors to accurately predict the future cash flow and operating cash flow of the company. Buy in excess of the expected earnings of the company, to avoid the less profitable company, is the investment for A long time, but in the a-share market, profit easily manipulated, uncertainty, but the operating activities cash flow is not easy to fabricate, and the index to reflect the production and operation of the enterprise actual situation, so this paper choose to use A machine learning algorithm to predict the indicators, and then predict the future development trend of the enterprise. Operating cash flow is defined as the income earned by an enterprise from providing various services or selling products in its production and operation activities, which is related to the survival and development of the enterprise and of great significance to the enterprise.This paper first introduces the basic idea, main methods, latest development and application of machine learning. And then introduces the framework of machine learning to predict business cash flow, this paper explores a variety of methods, including LR, XGBOOSTOOST, DNN algorithms, and bagging and stacking combination method. In the process of experiment, using the based on gradient descent algorithm, the effect of the sliding window algorithm optimization, based on the Max - min, the standard methods of the characteristics of the normal methods such as pretreatment, finally achieved a good prediction effect.