The Impact of Abnormal News Sentiment on Financial Markets

Stock portfolio selection using learning-to-rank algorithms with news sentiment

“…First, learning-to-rank algorithms produce accurate predictions of expected return rankings, and the proposed stock selection approach works robustly under different financial market conditions. Secondly, the sentiment indicators [31,38] [1] documented that the number of posts has significant correlation with market return. Second, financial news indicates market volatility level.…”

“…This trading strategy is based on the hypothesis that ranking of news sentiment reflects expected returns during the near future. In this study, we use the sentiment shock and trend indicators introduced in previous studies [31,38] to develop stock selection rules of holding long positions of the top 25% stocks and short positions of the bottom 25% stocks according to the stock rankings produced by learning-to-rank algorithms. In the experiments of portfolio strategies, we apply ListNet and RankNet in stock selection processes and test long-only and long-short strategies.…”

“…Second, financial news indicates market volatility level. In a previous study, we used linear regression to demonstrate that abnormal news sentiment has significant predictive power of implied volatility of S&P 500 index [38]. Findings of news impacts on financial markets led to further studies of sentiment-based algorithmic strategies.…”

News media analytics in finance: a survey

“…Vector autoregressive (VAR) models with multiple lags and joint significance tests go some way to alleviate aggregation bias, in that they reduce the temporal invariance from the entire formation period to the length of each lag and capture the linear dynamics between lags. Other studies explicitly use changes in temporal characteristics, such as sentiment spikes, momentum and trend changes, as news events (Uhl et al ., ; Yang et al ., ; Nooijen and Broda, ).…”

“…This returned 276 items, to which an additional 10 items (Wüthrich et al ., ; Chan, ; Kothari et al ., ; Mitra et al ., ; Tetlock, ; Leinweber and Sisk, ; Ferguson et al ., ; Yang et al ., ; Sinha, ; Manela and Moreira, ) were added from other sources, such as Google Scholar. To remove irrelevant items and establish a refined corpus subject to individual written review, the 286 initial items were filtered by removing those that:…”