An End-to-End Stock Recommendation Algorithm Study Based on Time-Frequency Consistency

The volatility and complexity of stock prices in the financial market make precise trend prediction a formidable challenge. Traditional stock prediction approaches often rely solely on either time-domain or frequency-domain information, which limits their ability to fully capture the multi-scale dynamics of stock prices, resulting in suboptimal prediction accuracy. To overcome these limitations, this paper presents an end-to-end stock recommendation algorithm grounded in time-frequency consistency. First, we introduce a time-frequency consistency analysis method that extracts both time-domain and frequency-domain features of stock prices concurrently, offering a more holistic view of trend fluctuations. Next, by applying prompt learning strategies, the model leverages pre-set prompts to identify optimal low-risk buying points within targeted time intervals, enhancing the decision-making process for stock recommendations. Finally, end-to-end model training facilitates seamless integration and automation from data input to stock recommendation output, enabling a fully streamlined prediction workflow. Experimental results indicate that this method surpasses traditional approaches in prediction accuracy and risk control, providing more dependable support for investor decisions.