Can AI Agent Be the New Assistant to Robo-advisers?

Overall Logic: The first step is data collection and cleaning; the second step involves experimenting with models for different data types to obtain trading signals; the third step is backtesting the trading strategies. The details are as follows:Part One: Sentiment-based Stock Selection1. First, fine-tune the model using some publicly available datasets from Xueqiu .2. Next, collect and clean actual news data.3. Perform sentiment analysis on the actual news data to obtain sentiment trading signals.4. Finally, backtest the trading strategy.Part Two: Predictive Stock SelectionThis part is divided into GLM (Generalized Linear Model) stock selection, ResNet stock selection, and their combination, used to obtain predictive trading signals, followed by backtesting based on different signals.Part Three: Indicator-based Stock Selection Construct composite indicators using GLM to generate trading signals from traditional financial data and backtest based on these signals.Part Four: Comprehensive Stock SelectionIntegrate the results from sentiment, predictive, and indicator-based stock selections for further filtering to reduce risk, and backtest based on these signals.