Investment Behaviors Can Tell What Inside: Exploring Stock Intrinsic Properties for Stock Trend Prediction

pp. 2376-2384, 2019.

Cited by: 3|Bibtex|Views42|DOI:https://doi.org/10.1145/3292500.3330663
EI
Other Links: dl.acm.org|dblp.uni-trier.de|academic.microsoft.com
Weibo:
We propose to take into account stock intrinsic properties in stock prediction tasks in order to enhance existing model based on dynamic inputs

Abstract:

Stock trend prediction, aiming at predicting future price trend of stocks, plays a key role in seeking maximized profit from the stock investment. Recent years have witnessed increasing efforts in applying machine learning techniques, especially deep learning, to pursue more promising stock prediction. While deep learning has given rise t...More

Code:

Data:

0
Introduction
  • Among a myriad of investment channels, the stock market has been continually considered as a big profitable potential.
  • Most of the traditional efforts on stock prediction rely on time-series analysis models, such as Autoregressive models [17], Kalman Filters [3], technical analysis [6], etc
  • These solutions create dynamic stock indicators, based on stock prices and volumes, as stochastic inputs and take the historical data of indicators to fit the stochastic trends.
  • With recent rapid development of deep learning, deep neural networks, especially recurrent neural networks (RNN), have been introduced as a promising substitute since its ability to model the sequential nature and non-linear structure within the stock prediction task [2, 10, 26, 31]
Highlights
  • Among a myriad of investment channels, the stock market has been continually considered as a big profitable potential
  • We develop a novel deep learning framework to integrate static stock intrinsic properties into the dynamic stock prediction task by modeling dynamic market state/trend
  • The recent efforts on stock trend prediction focused on leveraging deep neural networks with dynamic inputs in terms of price and volume indicators
  • To examine the quality of stock representations learned from mutual fund portfolios, we take some qualitative analysis to assess if the learned stock representations can capture the intrinsic properties
  • Most of the stocks in Livestock and Agriculture industries are clustered together into the third cluster. Such clustering results can clearly indicate that the stock representations extracted from mutual fund portfolios can carry certain intrinsic properties
  • We propose to take into account stock intrinsic properties in stock prediction tasks in order to enhance existing model based on dynamic inputs
Methods
  • The recent efforts on stock trend prediction focused on leveraging deep neural networks with dynamic inputs in terms of price and volume indicators.
  • The authors take advantage of LSTM and SFM as the RNN module, as shown in Figure 3, to model dynamic inputs.
  • Note that, among these state-of-the-art stock prediction approaches, none of them, to the best knowledge, ever leveraged the intrinsic properties extracted from mutual fund portfolios.
  • This method can be viewed as a specification of the method introduced in Section 3.1.2
Results
  • 5.1 Learned Stock Representations

    To examine the quality of stock representations learned from mutual fund portfolios, the authors take some qualitative analysis to assess if the learned stock representations can capture the intrinsic properties.
  • Table 3 shows three examples of obtained stock clusters in the second half year of 2015.
  • From this table, the authors can find that all the stocks in the first cluster belong to the basic industry while those in the second cluster are much related to the light industry.
  • Most of the stocks in Livestock and Agriculture industries are clustered together into the third cluster
  • Such clustering results can clearly indicate that the stock representations extracted from mutual fund portfolios can carry certain intrinsic properties
Conclusion
  • The authors propose to take into account stock intrinsic properties in stock prediction tasks in order to enhance existing model based on dynamic inputs.
  • The authors propose to extract stock intrinsic properties from mutual fund portfolios.
  • The authors develop a novel model to use static stock properties in a dynamic way by measuring the correlation between the market and the stock.
  • The authors plan to seek stock intrinsic properties from other valuable data and extend market state model in a dedicated way.
  • The authors will explore more useful investment behaviors of fund managers to improve stock prediction models
Summary
  • Introduction:

    Among a myriad of investment channels, the stock market has been continually considered as a big profitable potential.
  • Most of the traditional efforts on stock prediction rely on time-series analysis models, such as Autoregressive models [17], Kalman Filters [3], technical analysis [6], etc
  • These solutions create dynamic stock indicators, based on stock prices and volumes, as stochastic inputs and take the historical data of indicators to fit the stochastic trends.
  • With recent rapid development of deep learning, deep neural networks, especially recurrent neural networks (RNN), have been introduced as a promising substitute since its ability to model the sequential nature and non-linear structure within the stock prediction task [2, 10, 26, 31]
  • Methods:

    The recent efforts on stock trend prediction focused on leveraging deep neural networks with dynamic inputs in terms of price and volume indicators.
  • The authors take advantage of LSTM and SFM as the RNN module, as shown in Figure 3, to model dynamic inputs.
  • Note that, among these state-of-the-art stock prediction approaches, none of them, to the best knowledge, ever leveraged the intrinsic properties extracted from mutual fund portfolios.
  • This method can be viewed as a specification of the method introduced in Section 3.1.2
  • Results:

    5.1 Learned Stock Representations

    To examine the quality of stock representations learned from mutual fund portfolios, the authors take some qualitative analysis to assess if the learned stock representations can capture the intrinsic properties.
  • Table 3 shows three examples of obtained stock clusters in the second half year of 2015.
  • From this table, the authors can find that all the stocks in the first cluster belong to the basic industry while those in the second cluster are much related to the light industry.
  • Most of the stocks in Livestock and Agriculture industries are clustered together into the third cluster
  • Such clustering results can clearly indicate that the stock representations extracted from mutual fund portfolios can carry certain intrinsic properties
  • Conclusion:

    The authors propose to take into account stock intrinsic properties in stock prediction tasks in order to enhance existing model based on dynamic inputs.
  • The authors propose to extract stock intrinsic properties from mutual fund portfolios.
  • The authors develop a novel model to use static stock properties in a dynamic way by measuring the correlation between the market and the stock.
  • The authors plan to seek stock intrinsic properties from other valuable data and extend market state model in a dedicated way.
  • The authors will explore more useful investment behaviors of fund managers to improve stock prediction models
Tables
  • Table1: An example of a mutual fund, named ‘ChinaAMC Growth Fund’ with code 000001 in Chinese stock market, and its sampled half-year portfolios
  • Table2: The statistics of mutual fund portfolios data
  • Table3: Three examples of stock clusters generated based on stock representation extracted from mutual fund portfolios
Download tables as Excel
Related work
  • Recent work on stock prediction relies on two kinds of information sources: indicators from stock price/volume data and text from news and social medias.

    Technical indicators [6], i.e., mathematical calculations based on historical price, volume, or (in the case of futures contracts) open interest information [20], are proposed by financial experts at first for the purpose of discovering trading patterns of dynamic indicators. One of the most widely used models in the pattern recognition is Autoregressive (AR) model for linear and stationary time-series [17]. However, the non-linear and non-stationary nature of stock prices limits the application of AR models. Hence, substantial studies attempted to apply non-linear models to capture the complex dynamic market trend. With the development of deep learning, more scientists make efforts to exploit deep neural network for financial prediction [1, 12, 14, 16, 23, 30]. To further model the long-term dependency in time series, recurrent neural networks (RNNs), especially Long Short-Term Memory (LSTM) network, have also been employed in stock prediction [2, 10, 26]. In most recent time, a new RNN named the State Frequency Memory (SFM) is proposed by Zhang et.al [31] to discover the multi-frequency trading patterns.
Funding
  • Chi Chen and Chunxiao Xing are supported by NSFC 91646202 and National Key R&D Program of China SQ2018YFB140235
Reference
  • Ayodele Ariyo Adebiyi, Aderemi Oluyinka Adewumi, and Charles Korede Ayo. 2014. Comparison of ARIMA and artificial neural networks models for stock price prediction. Journal of Applied Mathematics 2014 (2014).
    Google ScholarLocate open access versionFindings
  • Ryo Akita, Akira Yoshihara, Takashi Matsubara, and Kuniaki Uehara. 2016. Deep learning for stock prediction using numerical and textual information. In Computer and Information Science (ICIS), 2016 IEEE/ACIS 15th International Conference on. IEEE, 1–6.
    Google ScholarLocate open access versionFindings
  • Ranjeeta Bisoi and Pradipta K Dash. 2014. A hybrid evolutionary dynamic neural network for stock market trend analysis and prediction using unscented Kalman filter. Applied Soft Computing 19 (2014), 41–56.
    Google ScholarLocate open access versionFindings
  • Louis KC Chan, Hsiu-Lang Chen, and Josef Lakonishok. 2002. On mutual fund investment styles. The Review of Financial Studies 15, 5 (2002), 1407–1437.
    Google ScholarLocate open access versionFindings
  • Prasanna Chandra. 2017. Investment analysis and portfolio management. McGraw-
    Google ScholarFindings
  • Robert D Edwards, John Magee, and WH Charles Bassetti. 2007. Technical analysis of stock trends. CRC press.
    Google ScholarFindings
  • Eric G Falkenstein. 1996. Preferences for stock characteristics as revealed by mutual fund portfolio holdings. The Journal of Finance 51, 1 (1996), 111–135.
    Google ScholarLocate open access versionFindings
  • Thomas Fischer and Christopher Krauss. 201Deep learning with long shortterm memory networks for financial market predictions. European Journal of Operational Research 270, 2 (2018), 654–669.
    Google ScholarLocate open access versionFindings
  • Brendan J Frey and Delbert Dueck. 2007. Clustering by passing messages between data points. science 315, 5814 (2007), 972–976.
    Google ScholarLocate open access versionFindings
  • Qiyuan Gao. 2016. Stock market forecasting using recurrent neural network. Ph.D.
    Google ScholarFindings
  • Felix A Gers, Nicol N Schraudolph, and Jürgen Schmidhuber. 2002. Learning precise timing with LSTM recurrent networks. Journal of machine learning research 3, Aug (2002), 115–143.
    Google ScholarLocate open access versionFindings
  • Mustafa Göçken, Mehmet Özçalıcı, Aslı Boru, and Ayşe Tuğba Dosdoğru. 2016. Integrating metaheuristics and artificial neural networks for improved stock price prediction. Expert Systems with Applications 44 (2016), 320–331.
    Google ScholarLocate open access versionFindings
  • Zura Kakushadze. 2016. 101 formulaic alphas. Wilmott 2016, 84 (2016), 72–81.
    Google ScholarLocate open access versionFindings
  • Kyoung-Jae Kim and Hyunchul Ahn. 2012. Simultaneous optimization of artificial neural networks for financial forecasting. Applied Intelligence 36, 4 (2012), 887–898.
    Google ScholarLocate open access versionFindings
  • Yehuda Koren, Robert Bell, and Chris Volinsky. 2009. Matrix factorization techniques for recommender systems. Computer 42, 8 (2009).
    Google ScholarLocate open access versionFindings
  • Leonel A Laboissiere, Ricardo AS Fernandes, and Guilherme G Lage. 2015. Maximum and minimum stock price forecasting of Brazilian power distribution companies based on artificial neural networks. Applied Soft Computing 35 (2015), 66–74.
    Google ScholarLocate open access versionFindings
  • Lili Li, Shan Leng, Jun Yang, and Mei Yu. 2016. Stock Market Autoregressive Dynamics: A Multinational Comparative Study with Quantile Regression. Mathematical Problems in Engineering 2016 (2016).
    Google ScholarFindings
  • Tie-Yan Liu et al. 2009. Learning to rank for information retrieval. Foundations and Trends® in Information Retrieval 3, 3 (2009), 225–331.
    Google ScholarLocate open access versionFindings
  • LR Medsker and LC Jain. 2001. Recurrent neural networks. Design and Applications 5 (2001).
    Google ScholarLocate open access versionFindings
  • John J Murphy. 1999. Technical analysis of the financial markets: A comprehensive guide to trading methods and applications. Penguin.
    Google ScholarFindings
  • Arman Khadjeh Nassirtoussi, Saeed Aghabozorgi, Teh Ying Wah, and David Chek Ling Ngo. 2015. Text mining of news-headlines for FOREX market prediction: A Multi-layer Dimension Reduction Algorithm with semantics and sentiment. Expert Systems with Applications 42, 1 (2015), 306–324.
    Google ScholarLocate open access versionFindings
  • David MQ Nelson, Adriano CM Pereira, and Renato A de Oliveira. 2017. Stock market’s price movement prediction with LSTM neural networks. In Neural Networks (IJCNN), 2017 International Joint Conference on. IEEE, 1419–1426.
    Google ScholarLocate open access versionFindings
  • Jigar Patel, Sahil Shah, Priyank Thakkar, and K Kotecha. 2015. Predicting stock market index using fusion of machine learning techniques. Expert Systems with Applications 42, 4 (2015), 2162–2172.
    Google ScholarLocate open access versionFindings
  • V Paul Pauca, Farial Shahnaz, Michael W Berry, and Robert J Plemmons. 2004. Text mining using non-negative matrix factorizations. In Proceedings of the 2004 SIAM International Conference on Data Mining. SIAM, 452–456.
    Google ScholarLocate open access versionFindings
  • G Preethi and B Santhi. 2012. STOCK MARKET FORECASTING TECHNIQUES: A SURVEY. Journal of Theoretical & Applied Information Technology 46, 1 (2012).
    Google ScholarLocate open access versionFindings
  • Akhter Mohiuddin Rather, Arun Agarwal, and VN Sastry. 2015. Recurrent neural network and a hybrid model for prediction of stock returns. Expert Systems with Applications 42, 6 (2015), 3234–3241.
    Google ScholarLocate open access versionFindings
  • Eberhard Schöneburg. 1990. Stock price prediction using neural networks: A project report. Neurocomputing 2, 1 (1990), 17–27.
    Google ScholarLocate open access versionFindings
  • Amnon Shashua and Tamir Hazan. 2005. Non-negative tensor factorization with applications to statistics and computer vision. In Proceedings of the 22nd international conference on Machine learning. ACM, 792–799.
    Google ScholarLocate open access versionFindings
  • Jianfeng Si, Arjun Mukherjee, Bing Liu, Qing Li, Huayi Li, and Xiaotie Deng. 2013. Exploiting topic based twitter sentiment for stock prediction. In Proceedings of the 51st Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers), Vol. 2. 24–29.
    Google ScholarLocate open access versionFindings
  • Jonathan L Ticknor. 2013. A Bayesian regularized artificial neural network for stock market forecasting. Expert Systems with Applications 40, 14 (2013), 5501–5506.
    Google ScholarLocate open access versionFindings
  • Liheng Zhang, Charu Aggarwal, and Guo-Jun Qi. 2017. Stock Price Prediction via Discovering Multi-Frequency Trading Patterns. In Proceedings of the 23rd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining.
    Google ScholarLocate open access versionFindings
  • Zhenkun Zhou, Jichang Zhao, and Ke Xu. 2016. Can online emotions predict the stock market in china?. In International Conference on Web Information Systems Engineering. Springer, 328–342.
    Google ScholarLocate open access versionFindings
Full Text
Your rating :
0

 

Tags
Comments