This paper presents a hybrid deep learning framework that combines Convolutional Neural Networks (CNN) and Gated Recurrent Units (GRU) for analyzing stock market sentiment from network text data and generating risk alerts. The CNN component is employed to preprocess and extract local features and patterns from large volumes of textual data, such as news articles or social media posts. These extracted feature sequences are then fed into the GRU model, which captures the temporal progression of emotional states and their potential impact on future market sentiment and risk. By leveraging CNN's strength in spatial feature extraction and GRU's ability to handle sequential dependencies, the proposed method effectively addresses the order dependence and long-term dependencies inherent in time series data. The integrated approach results in a detailed analysis of stock market sentiment and provides effective early warnings of future risks, demonstrating its potential for real-world financial risk management applications.