A Novel Approach to Economic Sentiment Index Based on Structured and Unstructured Data

This study proposes a new approach to analysing economic sentiment indexes and stock market movements: combining structured financial indicators with qualitative sentiment analysis and topic theme modelling in news texts. The study firstly points out the limitations of traditional economic analyses relying only on structured data. For example, the cyclicality of China's A-share market over the past decade - peaking in 2015 and then declining sharply - reflects market volatility that is difficult to fully capture with traditional methods. In this paper, five models are compared and it can be obtained that DNN is able to identify deep nonlinear patterns in data. Therefore it performs well in analysing both structured and unstructured data. In conclusion, this study highlights the shortcomings of traditional economic analyses that focus only on structured data and points out the importance of integrating unstructured data (e.g., market sentiment and news analyses) for a more accurate understanding of financial health and market dynamics.