Forecasting fund-related textual emotion trends on Weibo: A time series study

Abstract

Introduction: This paper reports a time series analysis of day-to-day emotional text related to fund investments on Weibo (Sina Corporation, Beijing, China). Methods: The present study employed web-crawler and text mining techniques through Python to obtain data from January 1, 2021 to December 31, 2021. Results: Using an auto-regressive integrated moving average model and vector auto-regressive model, the results indicated that fund performance was a significant predictor of fear, anger, and surprise expressions on Weibo. A relationship among emotions within a certain single fund was not found, but textual emotions could be predicted by ARIMA models within emotions. Discussion: The findings provide insight for media emotion analysis combining linguistic and temporal dimensions in both the communication and psychology disciplines.