Abstract:

The development of artificial intelligence and machine learning technology has greatly improved the efficiency of financial information processing， especially the ability to mine hidden data variables that have important impacts on the stock market. It has been found that investors’ attention has a great impact on the stock market. Therefore， it starts from the implicit variable of investors’ attention in this paper， and reversal online portfolio strategies are studied. Firstly， the code and abbreviation of the stock are used as search terms， and the sum of the two in Baidu Index is obtained through web crawler technology. Based on search index and smoothing method， a measure of investor attention is constructed. Then， combined with the reversal effect of the stock market， the reversal online portfolio strategy is designed which takes into account the investors’ attention. Finally， a backtest is conducted using six Chinese stock datasets with differences in data sources and stock sizes. The results indicate that the strategy designed in this paper performs well in terms of returns and risks， improving existing online portfolio strategies based on mean reversal effect. The BD-AC strategy outperforms benchmark and traditional strategies on most datasets. Furthermore， sensitivity analysis is also conducted on two parameters of the strategy， including the time window and the lag window， which demonstrates good robustness within a certain range of parameters. It contributes to the field of online portfolio research in this paper by constructing a measure of investor attention based on Baidu index and introducing it into the Anticor model to improve the strategy.

Key words: investors’ attention, Baidu index, mean reversal effect, online portfolio