Research on Investment Bank Risk Monitoring and Early Warning Model Combining Factor Analysis and Artificial Neural Network

As artificial intelligence and big data continue to grow, quantitative trading has become more popular, encouraging investors to develop stronger strategies for better returns. An effective investment strategy should be adaptable, able to control risk, profitable, and easy to implement. Two key components of such strategies are "timing" and "stock selection." Combining these elements can lower risk and increase returns, helping markets operate more smoothly. This research provides an investment banking and warning model that combines time planning with stock options. Time planning: One strategy uses a model based on historical data - the Hidden Markov Model (HMM) - to forecast the market and guide trading decisions.Backtesting results show that the selected assets generate excess returns while keeping risk under control. The stock selection strategy uses a neural network that includes seven main factors (valuation, momentum, growth, technical, leverage, quality, and sentiment) to build a factor pool. After applying the information coefficient (IC) and principal component analysis (PCA), the data were incorporated into the BP neural network.A Particle Swarm Optimization (PSO) algorithm is used to fine-tune the network parameters. The combined strategy outperforms the individual strategies, offering better returns and risk management. This research confirms that combining these strategies is effective and offers valuable insights for future quantitative investment models.