A representation and classification method for collective investor attention in the financial market

Abstract

Introduction: It is increasingly becoming integral to analyze the collected information effectively.Methods: We propose a representation and classification method for collective investor attention in the financial market, taking the Chinese stock market as an example. The method includes three key steps: 1) converting the hourly search volume of each stock per week to an image representation for describing the changes of collective investor attention; 2) extracting features of each image by utilizing a self-encoding algorithm in deep learning; and 3) clustering generated images by K-means to arrange stocks into different groups.Results: The empirical results show that the portfolio considering the clustering information outperforms the HS300 index.Discussion: The method may not only use deep learning features for stock similarity measurement, but also shed some light on profoundly understanding the mechanisms of the collective investor attention for the financial market.