对股票市场的系统稳定性的研究是金融领域的热门问题。股票市场可以看作一个复杂的网络，网络的节点是由表示节点间相关性强度的边连接起来的股票。股票市场的非线性和随机性使得其宏观稳定性和市场波动的刻画和预测是一个困难的问题。本文基于离散Ollivier-Ricci曲率刻画中国股票市场网络的高阶非线性结构，利用该方法对2010-2022年的40支中国股市上证50指数成分股的5种指标进行了分析，并结合主成分分析，确定最适宜量化市场的主要指标。在此基础上，我们使用WD-LSTM混合模型来分析曲率时间序列，预测市场的未来趋势，并借助网络曲率，使用聚类算法给出股票市场板块的一种新的划分方式。我们的结果表明，基于成交量的Ollivier-Ricci曲率对衡量和预测市场稳定性具有较好的效果，可以作为判断未来国内股票市场风险和波动的一个主要指标，并且通过网络曲率可以给出股票市场合理的板块划分方式。

The research on the system stability of the stock market is a hot issue in the financial field. The stock market can be viewed as a complex network whose nodes are stocks connected by edges that indicate the strength of correlation between nodes. The nonlinearity and randomness of stock market make it a difficult problem to describe and predict its macro stability and market fluctuation. Based on discrete Ollivier-Ricci curvature, this paper describes the high-order nonlinear structure of China's stock market network, and uses this method to analyze five indexes of 40 Shanghai 50 Index stocks from 2010 to 2022. Combined with principal component analysis, the most suitable index for quantifying the market is determined. On this basis, we use the WD-LSTM hybrid model to analyze the curvature time series, predict the future trend of the market, and with the help of the network curvature, use the clustering algorithm to give a new division of the stock market sector. Our results show that the Ollivier-Ricci curvature based on volume has a good effect on measuring and predicting market stability, and can be used as a main index to judge the risk and volatility of the domestic stock market in the future. Moreover, the network curvature can give a reasonable way to partition the stocks of the market.