A novel hybrid stock selection method with stock prediction
Graphical abstract
Introduction
Stock selection has long been identified as an important but challenging topic in the research area of financial market analysis [1], [2]. Like other similar research questions (e.g., stock prediction and stock portfolio recommendation), it is also in an attempt to facilitate the decision-making task for investment [3]. Nevertheless, it holds its own unique characteristic, in terms of a different specific task and hence different techniques. In particular, stock selection tends to distinguish ‘good’ stocks from ‘bad’ stocks, based on evaluation or scoring models [4], [5]; stock prediction reveals future information for stock prices and trends, based on forecasting models [6], [7], [8]; and stock portfolio recommendation makes optimal distributions of investment among stocks, based on optimization models [9], [10]. Accordingly, stock selection could serve as a preliminary step for stock portfolio recommendations to provide target stocks deserving study; and stock prediction could facilitate stock selection to capture the stock markets in the future [3], [11]. Therefore, this study especially focuses on stock selection, and introduce stock prediction for model improvements.
The core part of stock selection is the stock scoring mechanism for evaluating the value of a stock, which is the basis on which top-valued stocks (corresponding to high potential returns and low potential risk levels) are selected [12]. According to the existing literature, linear models, which describe the score of a stock as the sum of a set of weighted factors, are the predominantly used stock scoring models [13]. Accordingly, factor determination is an essential part of stock scoring; popular factors include yield factors, liquidity factors, risk factors, growth factors, momentum factors, etc., which effectively reflect the history and/or current features of a stock [2], [4], [12]. However, such information may lose efficacy in capturing the time-varying features of the associated stocks in the future, due to the extreme complexity and volatility of stock markets [14]. Furthermore, the success of stock selection is majorly contingent upon the future performance of stock markets; however, traditional stock selection models that consider only history and current information may have difficulty describing the future volatility of stock markets. Thus, the introduction of factors regarding future features of stock markets offers a new perspective for improving stock selection.
To explore the future features of stock markets, various forecasting algorithms have been employed, of which, computational intelligence (CI) (or artificial intelligence (AI)) has become increasingly dominant due to its powerful learning capability and high prediction accuracy. Typical CI techniques in stock market prediction (for stock prices, stock returns, market indexes, etc.) are artificial neural networks (ANNs) [15], [16], [17], [18] and support vector machines (SVMs) [19], [20], [21], [22]. The existing studies have investigated their effectiveness in providing valid information about future features of stock markets, which could in turn serve stock evaluation and selection [3], [23]. Among various CI learning paradigms, an emerging ANN (i.e., extreme learning machine (ELM), which effectively tackles the problems of time consumption and local optima in traditional CI models by using randomly fixed parameters) shows clear superiority in financial market prediction in terms of much higher estimation accuracy and much faster computing speed [24], [25], [26]. Therefore, we implement the efficient CI tool in stock market prediction and further improve stock selection by incorporating predicted future features.
This paper proposes a novel hybrid stock selection method that incorporates stock prediction to effectively capture the future features of complex stock markets. The proposed method has two main steps: stock prediction and stock scoring. First, stock returns for the next period are predicted based on the emerging CI technique of ELM, which boasts fast computing speed and good generalization performance [27]. Second, the predicted factor and various fundamental factors (popular in existing models) are introduced into a typical linear stock scoring mechanism to evaluate the value of each candidate stock; highly valued stocks are selected to formulate an equally weighted portfolio. In the stock scoring mechanism, the weight terms of different factors are optimized using a typical CI programming technique, i.e., differential evolution (DE), which has a simple design but efficient performance [28]. Compared with existing stock selection models, our novel method makes the following contributions: (1) this method might be the first attempt to couple stock prediction with stock scoring, forming a novel stock selection method with the predicted factor regarding the future features of stock markets; (2) the proposed model is applied to the A-share market of China to verify its effectiveness and validity in stock selection and portfolio formulation, and typical stock selection models (without stock selection) and similar counterparts (with other forecasting models, factor designs, optimization algorithms or fitness functions) are also conducted for comparison.
The main aim of this paper is to propose a novel hybrid stock selection model with stock prediction for capturing the future features of stock markets and to verify its superiority over similar counterparts. The rest of this paper is organized as follows. Section 2 formulates the novel methodology in detail. Section 3 reports the empirical results and discusses the effectiveness of the proposed model. Section 4 concludes the paper and outlines potential directions for future research.
Section snippets
Literature review
Stock selection has become an increasingly hot issue in the field of finance research, as recent interesting studies listed in Table 1.
On one hand, diverse methods have been applied to stock selection, which fall into two main categories, i.e., statistical methods (e.g., linear regression [11], fuzzy analysis [33], cluster analysis [34], and ordered weighted averaging [29], [36]) and CI techniques (e.g., ANN [11], [30], [32], SVM [5], [30], DE [4], [31], GA [4], PSO [4]). Of them, the CI
Methodology formulation
A novel hybrid stock selection method with stock prediction is formulated in this section. Section 3.1 presents the general framework of the novel model. Sections 3.2 Stock prediction, 3.3 Stock scoring elaborate on the two major steps, i.e., stock prediction and stock scoring, respectively. Section 3.4 shows a toy example of the proposed stock selection model.
Empirical study
For illustration and verification, the novel hybrid stock selection method is applied for portfolio construction in the A-share market of China. Section 4.1 designs the empirical study, Section 4.2 investigates the effectiveness of the proposed model relative to average market performance and A-share index, Section 4.3 discusses the superiority of the proposed model over various typical stock selection models, and Section 4.4 summarizes the empirical study.
Conclusions
This paper proposes a novel hybrid stock selection model by incorporating stock prediction to effectively capture the future features of complex stock markets. The novel model majorly contributes to the literature from two main perspectives. First, it might be the first attempt to couple stock prediction with stock selection to form a novel stock selection method with a predicted factor capturing future stock markets. Second, it is applied to the A-share market of China and compared with
References (51)
A hybrid stock selection model using genetic algorithms and support vector regression
Appl. Soft. Comput.
(2012)- et al.
Surveying stock market forecasting techniques–part II: Soft computing methods
Expert Syst. Appl.
(2009) - et al.
Feature selection and parameter optimization of a fuzzy-based stock selection model using genetic algorithms
Int. J. Fuzzy Syst.
(2012) - et al.
Evidence that stock prices do not fully reflect the implications of current earnings for future earnings
J. Account. Econ.
(1990) - et al.
Application of artificial neural network for the prediction of stock market returns: The case of the Japanese stock market
Chaos Solitons Fractals
(2016) - et al.
A hybrid approach based on neural networks and genetic algorithms for detecting temporal patterns in stock markets
Appl. Soft. Comput.
(2007) - et al.
A new constructive neural network method for noise processing and its application on stock market prediction
Appl. Soft. Comput.
(2014) - et al.
News impact on stock price return via sentiment analysis
Knowl.-Based Syst.
(2014) - et al.
Improved v-support vector regression model based on variable selection and brain storm optimization for stock price forecasting
Appl. Soft. Comput.
(2016) - et al.
Support vector regression with chaos-based firefly algorithm for stock market price forecasting
Appl. Soft. Comput.
(2013)
Improving returns on stock investment through neural network selection
Expert Syst. Appl.
A self adaptive differential harmony search based optimized extreme learning machine for financial time series prediction
Swarm Evol. Comput.
Financial technical indicator based on chaotic bagging predictors for adaptive stock selection in Japanese and american markets
Physica A
Analysis of microarray data using Z score transformation
J. Mol. Diagn.
Extreme learning machine: Theory and applications
Neurocomputing
A randomized-algorithm-based decomposition-ensemble learning methodology for energy price forecasting
Energy
Stock selection: an innovative application of genetic programming methodology
Stock selection with a novel sigmoid-based mixed discrete-continuous differential evolution algorithm
IEEE Trans. Knowl. Data Eng.
Multi-factor stock selection model based on kernel support vector machine
J. Math. Res.
A topic-based sentiment analysis model to predict stock market price movement using Weibo mood
Web Intell.
Combining multiple algorithms for portfolio management using combinatorial fusion
Neural network copula portfolio optimization for exchange traded funds
Quant. Finance
Cited by (71)
A self-attention based cross-sectional return forecasting model with evidence from the Chinese market
2024, Finance Research LettersQuantitative stock portfolio optimization by multi-task learning risk and return
2024, Information FusionA generalization of multi-source fusion-based framework to stock selection
2024, Information FusionA novel hybrid model for stock price forecasting integrating Encoder Forest and Informer
2023, Expert Systems with ApplicationsMulti-agent model based proactive risk management for equity investment
2023, Engineering Applications of Artificial Intelligence