Reliability study of stock index forecasting in volatile and trending cities using public sentiment ——based on word2Vec and LSTM models

ABSTRACT

Accurate forecasting of the stock market has always been a key concern of academics and investors, and few researchers have discussed the reliability of stock index forecasting in volatile and trending markets based on public sentiment. This article will first use Word2Vec and CNN to classify the sentiment of 754,000 text data excavated from the Oriental Fortune Stock Forum and construct public sentiment indicators; then, select characteristic parameters such as the closing value of the Shanghai Composite Index, the inflow of northbound funds, and the exchange rate of RMB against the US dollar. Introducing sentiment indicators and building an LSTM model to explore the effect of public sentiment factors on the prediction of the Shanghai Composite Index in the unilateral rise, unilateral fall, and volatile markets. The research shows that the reliability of using public sentiment to predict unilateral falling and volatile markets is high, especially the prediction error of predicting volatile markets is the smallest. In addition, it is also found that due to the existence of the ‘disposition effect’, the error is significantly larger when using public sentiment factors to predict the unilateral rising market.

KEYWORDS:

• Long short-term memory
• Public emotion
• Stock index forecast
• Convolutional neural networks
• The Shanghai composite index

JEL CLASSIFICATION:

• C33
• G14

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This research is supported by the National Natural Science Foundation of China (No. 71701036) and the Fundamental Research Funds for the Central Universities (No.N2123022).