SSE forecasts based on market-sentiment dual anchoring.

Area of Science:

• * Financial econometrics and computational finance.
• * Machine learning applications in quantitative trading.
• * Behavioral finance and market psychology.

Background:

• * Anchoring bias is a significant cognitive effect in financial markets.
• * Existing forecasting models often fail to integrate market dynamics with investor sentiment effectively.
• * The Shanghai Stock Exchange (SSE) presents complex dynamics influenced by sentiment.

Purpose of the Study:

• * To develop a novel forecasting model integrating dual market-sentiment anchoring with advanced deep learning.
• * To enhance the precision and lead time of Shanghai Stock Exchange (SSE) index predictions.
• * To evaluate the efficacy of spatial feature extraction via CNN2D against temporal feature extraction via CNN1D.

Main Methods:

• * Development of a dual market-sentiment anchoring multivariate feature matrix using SSE index extremes and sentiment data.
• * Implementation of a Market Sentiment Dual Anchoring CNN2D-ABiLSTM (MSD-CNN2D-ABiLSTM) model.
• * Utilizing CNN2D for spatial feature extraction, BiLSTM for temporal feature integration, and an attention mechanism.

Main Results:

• * Achieved prediction accuracy exceeding 90% and R2 value >95% for 1-2 day ahead SSE index forecasting.
• * Demonstrated effective forecasting performance up to 10 trading days ahead.
• * CNN2D's local spatial feature extraction showed superior predictive advantage over CNN1D's short-term temporal features.

Conclusions:

• * The proposed MSD-CNN2D-ABiLSTM model offers a significant advancement in financial market forecasting.
• * Dual anchoring and deep learning integration provide robust and accurate SSE index predictions.
• * Spatial feature extraction is crucial for capturing complex market structures in forecasting.