In today’s financial markets, investors face high volatility and complexity, which makes strategic decision-making and risk management particularly challenging. To address this issue, this study introduces a financial prediction model based on Long Short Term Memory (LSTM) networks, aimed at effectively capturing time series features. Firstly, this paper uses LSTM neural network to train historical financial data. Next, this paper extracts features at different time scales through a multi-layer network structure to enhance the robustness of the model. Meanwhile, this paper introduces reinforcement learning strategies to optimize the decision-making process and achieve effective risk control. In the experimental conclusion, the RMSE (root mean square error) value of the LSTM based model in terms of prediction accuracy is 2.35. After incorporating dynamically adjusted risk control strategies, the initial capital increased from $10000 to $22785. In the above data conclusions, the financial modeling method based on LSTM proposed in this study is significantly superior to traditional methods in terms of accuracy and risk control ability in strategic decision-making, demonstrating its advantages in dealing with complex and dynamic changes in financial markets.