News volume information: Beyond earnings forecasting in a global stock selection model

A modular Takagi-Sugeno-Kang (TSK) system based on a modified hybrid soft clustering for stock selection

“…Therefore, TSK system can model the stock selection problem properly. The proposed RFNRC-TSK-MBE model could reach IC value of 18.15% on average which is much more than the other stock selection models in the literature [60,88,93,94]. The genetic programming model provided by Becker et al [60] could reach to a maximum IC of 8%.…”

“…The authors claimed that 9% is a high IC value and their models were successful at the stocks' ranking. Also, Gillam et al [93] studied on the earnings prediction in a global stock selection model. They could improve the predictability of the model to the IC of 6%.…”

“…implemented our proposed ensemble learning model on 150 Iranian companies with different activity sectors listed on Tehran Stock Exchange (TSE) to develop a modular TSK system for stock selection. Based on the experimental results, our developed modular system could appropriately select the stocks with information coefficient of 18.15% on average, which is a good performance with respect to the previous researches[60,88,93]. Furthermore, our proposed system significantly outperformed the single TSK system (IC=8.5%) and also other modular TSK systems, i.e., fuzzy noise rejection clustering based TSK system with averaging ensemble (IC=12.74%) as well as rough-fuzzy noise rejection clustering based TSK system with averaging ensemble (IC=15.24%).…”

The existence and persistence of financial anomalies: What have you done for me lately?

“…Now-classical financial anomalies, as identified in Dimson (1988), Jacobs and Levy (1988), and Levy (1999) exist and have persisted. Moreover, the recent findings of Gillam, Guerard Jr., and Cahan (2015) suggest that earnings transcripts, commonly available to investors and often reported in the news, contain information that offers statistical support for inclusion in the portfolio creation process. Alternative data and predictive analytics, see Kuhn & Johnson (2013), new data sources and modeling techniques, offer the potential for investor risk-adjusted return enhancement.…”

Comparison of Stock Selection Methods: An Empirical Research On The Borsa Istanbul

“…Methods for a stock selection to generate a successful portfolio vary from type to type, such as predicting future stock prices by studying their past patterns (Goumatianos et al, 2013), predicting risk by using size or book-to-price ratio as firm-specific characteristics (Lucas et al, 2002), comparing performances of previous stock selection methodologies based on liquidity, size, mean reversion, and momentum (van der Hart et al, 2003), applying learning-to-rank algorithms to understand investors' sentiment toward a group of stocks by comparing long-term and short-term performances (Song et al, 2017), using candlestick charts to predict future returns to generate a cherry-picked portfolio (Horton, 2009), using case-based reasoning (CBR) relying on fundamental and technical analyses to recognize winning stocks around earning announcements by comparing classification accuracy and Sharpe ratio (Ince, 2014), using false discovery rate (FDR) to examine the model selections to be used in stock selection (Cuthbertson & Nitzche, 2013), using abnormal news volume information and rate of analysts' attention toward targeted stocks to detect golden stocks for generating a global portfolio (Gillam et al, 2015), analyzing stock-buying or stockselling actions of mutual fund firms (Ratanabanchuen & Saengchote, 2020), using Markov decision process on genetic algorithms to define trading strategies (Chang & Lee, 2017), using fuzzy model based on fuzzy ranking (Tiryaki & Ahlatcioglu, 2005), using chaotic bagging indicator to select risk-averse actions to allocate stocks (Suzuki & Okhura, 2016), using consensus temporary earnings forecasts (CTEF) data (Xia et al, 2015), combining analysts' forecasts, momentum data, and fundamental ratios of firms into a model (Guerard et al, 2015), investigating large shareholders' behaviors toward listed stocks (Sun et al, 2020), and using Gordon model improved with multiple criteria decision making (MCDM) model (Lee et al, 2009). Thus, the purpose of all stock selection methods defined above-and similar methods based on statistical, linear regression, fuzzy analyses, cluster analyses, and weighted average stock selection (Yang et al, 2019), and the methods revealed in Section 2-is generating higher returns and providing positive-alpha situations at least in the portfolio management business.…”