Individualized Indicator for All: Stock-wise Technical Indicator Optimization with Stock Embedding

pp. 894-902, 2019.

Cited by: 1|Bibtex|Views61|DOI:https://doi.org/10.1145/3292500.3330833
EI
Other Links: dl.acm.org|dblp.uni-trier.de|academic.microsoft.com
Weibo:
With the lack of the information generated from stock embedding, there is still obvious gap compared with out proposed model, which implies the effectiveness of our TTIO framework

Abstract:

As one of the most important investing approaches, technical analysis attempts to forecast stock movement by interpreting the inner rules from historic price and volume data. To address the vital noisy nature of financial market, generic technical analysis develops technical trading indicators, as mathematical summarization of historic pr...More

Code:

Data:

0
Introduction
  • Technical analysis [15, 22, 24], as one of essential approaches in quantitative investment, focuses on interpreting and forecasting stock movements in terms of its price and volume.
  • The central assumption in technical analysis lies in that all relevant information for investment decision is reflected by price and volume movement.
  • Fundamental analysis involves analyzing a company’s financial statements to determine the fair value of the business, while technical analysis assumes that a stock’s price already reflects all publicly-available information and instead focuses on the statistical analysis.
  • The price and volume data is comprised of all the information to make prediction in technical analysis
Highlights
  • Technical analysis [15, 22, 24], as one of essential approaches in quantitative investment, focuses on interpreting and forecasting stock movements in terms of its price and volume
  • The Complex seeks to combine the information of technical indicators and stock embedding with a multi-layer neural network and learns a new indicator one-to-one instead of scaling based on the original indicators
  • With the lack of the information generated from stock embedding, there is still obvious gap compared with out proposed model, which implies the effectiveness of our TTIO framework
  • We propose a general and explainable framework to optimize technical indicator with hidden knowledge mined from external resources
  • We propose a novel idea to leverage the difference in terms of indicator’s stock-wise affinity and take a data mining view to learn the stock representation, by mining knowledge repository from collective behaviors of experienced investors
  • The results show that our optimized indicator representation are significantly better and more stable on recent years
  • The indicators generated through our model do not give temporal difference we adapt it to the real world with a rough rotation learning method
Methods
  • To evaluate the performance of the approach, the authors design four baselines.

    Raw is the raw indicators generated based on the mathematical calculation on stocks’ price or volume.

    Norm re-scales the raw technical indicators by directly do the standard normalization calculations.
  • To evaluate the performance of the approach, the authors design four baselines.
  • Raw is the raw indicators generated based on the mathematical calculation on stocks’ price or volume.
  • Norm re-scales the raw technical indicators by directly do the standard normalization calculations.
  • Comparing with this approach will show how good enough simple scaling (a) 2016 (b) 2015 (c) 2014
Results
  • The Complex seeks to combine the information of technical indicators and stock embedding with a multi-layer neural network and learns a new indicator one-to-one instead of scaling based on the original indicators.
  • This method results of the over-fitting problem and represents low IC.
  • With the lack of the information generated from stock embedding, there is still obvious gap compared with out proposed model, which implies the effectiveness of the TTIO framework
Conclusion
  • The authors propose a general and explainable framework to optimize technical indicator with hidden knowledge mined from external resources.
  • The authors propose a novel idea to leverage the difference in terms of indicator’s stock-wise affinity and take a data mining view to learn the stock representation, by mining knowledge repository from collective behaviors of experienced investors.
  • The indicators generated through the model do not give temporal difference the authors adapt it to the real world with a rough rotation learning method.
  • Dynamically optimizing technical indicators will be left for the future work
Summary
  • Introduction:

    Technical analysis [15, 22, 24], as one of essential approaches in quantitative investment, focuses on interpreting and forecasting stock movements in terms of its price and volume.
  • The central assumption in technical analysis lies in that all relevant information for investment decision is reflected by price and volume movement.
  • Fundamental analysis involves analyzing a company’s financial statements to determine the fair value of the business, while technical analysis assumes that a stock’s price already reflects all publicly-available information and instead focuses on the statistical analysis.
  • The price and volume data is comprised of all the information to make prediction in technical analysis
  • Objectives:

    The authors aim to obtain effective representations to reflect stock properties based on the rule that stocks with similar properties have similar representations.
  • Methods:

    To evaluate the performance of the approach, the authors design four baselines.

    Raw is the raw indicators generated based on the mathematical calculation on stocks’ price or volume.

    Norm re-scales the raw technical indicators by directly do the standard normalization calculations.
  • To evaluate the performance of the approach, the authors design four baselines.
  • Raw is the raw indicators generated based on the mathematical calculation on stocks’ price or volume.
  • Norm re-scales the raw technical indicators by directly do the standard normalization calculations.
  • Comparing with this approach will show how good enough simple scaling (a) 2016 (b) 2015 (c) 2014
  • Results:

    The Complex seeks to combine the information of technical indicators and stock embedding with a multi-layer neural network and learns a new indicator one-to-one instead of scaling based on the original indicators.
  • This method results of the over-fitting problem and represents low IC.
  • With the lack of the information generated from stock embedding, there is still obvious gap compared with out proposed model, which implies the effectiveness of the TTIO framework
  • Conclusion:

    The authors propose a general and explainable framework to optimize technical indicator with hidden knowledge mined from external resources.
  • The authors propose a novel idea to leverage the difference in terms of indicator’s stock-wise affinity and take a data mining view to learn the stock representation, by mining knowledge repository from collective behaviors of experienced investors.
  • The indicators generated through the model do not give temporal difference the authors adapt it to the real world with a rough rotation learning method.
  • Dynamically optimizing technical indicators will be left for the future work
Tables
  • Table1: A set of popular technical indicators with respective calculation formulas
  • Table2: Top 5 stocks that obtained the maximum and minimum raw scaling weight on Bias, MACD and ROC
Download tables as Excel
Related work
  • Technical indicator is the fundamental tool in technical analysis. Previous work on indicator optimization can be roughly divided into two classes: hard-crafted indicator and indicator by deep learning.

    Hand-crafted indicator has been proposed by experienced investors and economists decades ago[9, 31, 34, 37]. Gunasekarage et al [11] analyzed the performance of the Simple Moving Average(SMA) indicator using index data for four emerging South Asian capital markets, and proved its predictive ability to generate excess returns. Chong and Ng [25] reported that the Relative Strength Index(RSI) as well as Moving Average ConvergenceDivergence(MACD), can generate return higher than buy-and-hold strategy in most cases. Instead of considering on indicators that limit to one stock, Gatev et al [8] tested a hedge fund equity trading strategy concerning the distance and the correlation of the prices. The experiments showed that trading suitably formed pairs of stocks exhibited profits, which were robust to conservative estimates of transaction costs.
Reference
  • Steven B Achelis. 200Technical Analysis from A to Z. McGraw Hill New York.
    Google ScholarFindings
  • Jacob Benesty, Jingdong Chen, Yiteng Huang, and Israel Cohen. 2009. Pearson correlation coefficient. In Noise reduction in speech processing. Springer, 1–4.
    Google ScholarFindings
  • LÃľon Bottou. 1998. Online Algorithms and Stochastic Approximations.
    Google ScholarFindings
  • Shaosheng Cao, Wei Lu, and Qiongkai Xu. 2016. Deep neural networks for learning graph representations. In Thirtieth AAAI Conference on Artificial Intelligence. 1145–1152.
    Google ScholarLocate open access versionFindings
  • Rodolfo C. Cavalcante, Rodrigo C. Brasileiro, Victor L. F. Souza, Jarley P. Nobrega, and Adriano L. I. Oliveira. 2016. Computational Intelligence and Financial Markets: A Survey and Future Directions. Expert Systems with Applications 55, 1 (2016), 194–211.
    Google ScholarLocate open access versionFindings
  • Chalothon Chootong and Ohm Sornil. 2012. Trading Signal Generation Using A Combination of Chart Patterns and Indicators. International Journal of Computer Science Issues 9, 6 (2012).
    Google ScholarLocate open access versionFindings
  • Y. Deng, F. Bao, Y. Kong, Z. Ren, and Q. Dai. 201Deep Direct Reinforcement Learning for Financial Signal Representation and Trading. IEEE Transactions on Neural Networks and Learning Systems 28, 3 (2017), 653.
    Google ScholarLocate open access versionFindings
  • Evan Gatev, William N. Goetzmann, and K. Geert Rouwenhorst. 2006. Pairs Trading: Performance of a Relative-Value Arbitrage Rule. The Review of Financial Studies 19, 3 (2006), 797–827.
    Google ScholarLocate open access versionFindings
  • Janos Gertler and Hong Shun Chang. 1986. An instability indicator for expert control. IEEE Control Systems Magazine 6, 4 (1986), 14–17.
    Google ScholarLocate open access versionFindings
  • Aditya Grover and Jure Leskovec. 2016. node2vec: Scalable Feature Learning for Networks. 2016 (2016), 855–864.
    Google ScholarFindings
  • Abeyratna Gunasekarage and David M Power. 2001. The profitability of moving average trading rules in South Asian stock markets. Emerging Markets Review 2, 1 (2001), 17–33.
    Google ScholarLocate open access versionFindings
  • However. 2014. Comparison of ARIMA and Artificial Neural Networks Models for Stock Price Prediction. Journal of Applied Mathematics,2014,(2014-3-19) 2014, 1 (2014), 1–7.
    Google ScholarLocate open access versionFindings
  • Why Read It. 1959. Portfolio selection: efficient diversification of investments.
    Google ScholarFindings
  • Kyoung-jae Kim. 2003. Financial time series forecasting using support vector machines. Neurocomputing 55, 1-2 (2003), 307–319.
    Google ScholarLocate open access versionFindings
  • Charles Kirkpatrick. 2006. Technical analysis: the complete resource for financial market technicians. FT Press.
    Google ScholarFindings
  • Muneesh Kumar and Sanjay Sehgal. 2004. Company Characteristics and Common Stock Returns: The Indian Experience. Vision 8, 2 (2004), 33–45.
    Google ScholarLocate open access versionFindings
  • Andy Liaw, Matthew Wiener, et al. 2002. Classification and regression by randomForest. R news 2, 3 (2002), 18–22.
    Google ScholarFindings
  • Laurens van der Maaten and Geoffrey Hinton. 2008. Visualizing data using t-SNE. Journal of machine learning research 9, Nov (2008), 2579–2605.
    Google ScholarLocate open access versionFindings
  • Massoud Metghalchi, Juri Marcucci, and Yung-Ho Chang. 2012. Are moving average trading rules profitable? Evidence from the European stock markets. Applied Economics 44, 12 (2012), 1539–1559.
    Google ScholarLocate open access versionFindings
  • Tomas Mikolov, Kai Chen, Greg Corrado, and Jeffrey Dean. 2013. Efficient Estimation of Word Representations in Vector Space. Computer Science (2013).
    Google ScholarLocate open access versionFindings
  • Tomas Mikolov, Ilya Sutskever, Kai Chen, Greg Corrado, and Jeffrey Dean. 2013. Distributed representations of words and phrases and their compositionality. 26 (2013), 3111–3119.
    Google ScholarFindings
  • Christopher J Neely et al. 1997. Technical analysis in the foreign exchange market: a layman’s guide. Federal Reserve Bank of St. Louis Review Sep (1997), 23–38.
    Google ScholarLocate open access versionFindings
  • Christopher J Neely, David E Rapach, Jun Tu, and Guofu Zhou. 2011. Forecasting the Equity Risk Premium: The Role of Technical Indicators. Social Science Electronic Publishing 60, 7 (2011), 1772–1791.
    Google ScholarLocate open access versionFindings
  • Christopher J Neely and Paul A Weller. 1997. Technical Analysis in the Foreign Exchange Market. Social Science Electronic Publishing 79, No. 2011-001 (1997), 343–373.
    Google ScholarLocate open access versionFindings
  • Wing Kam Ng. 2008. Technical analysis and the London stock exchange: testing the MACD and RSI rules using the FT30. Applied Economics Letters 15, 14 (2008), 1111–1114.
    Google ScholarLocate open access versionFindings
  • Mathias Niepert, Mohamed Ahmed, and Konstantin Kutzkov. 2016. Learning convolutional neural networks for graphs. In International Conference on International Conference on Machine Learning. 2014–2023.
    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 An International Journal 42, 4 (2015), 2162–2172.
    Google ScholarLocate open access versionFindings
  • Bryan Perozzi, Rami Alrfou, and Steven Skiena. 2014. DeepWalk: online learning of social representations. (2014), 701–710.
    Google ScholarFindings
  • Martin J Pring. 2002. Technical analysis explained. McGraw-Hill Companies.
    Google ScholarFindings
  • Qin Qin, Qing Guo Wang, Jin Li, and Shuzhi Sam Ge. 2013. Linear and Nonlinear Trading Models with Gradient Boosted Random Forests and Application to Singapore Stock Market. Journal of Intelligent Learning Systems and Applications 05, 1 (2013), 1–10.
    Google ScholarLocate open access versionFindings
  • Alejandro RodrÃŋguez-GonzÃąlez, ÃĄngel GarcÃŋa-Crespo, Ricardo Colomo-Palacios, Fernando GuldrÃŋs Iglesias, and Juan Miguel GÃşmez-BerbÃŋs. 2011. CAST: Using neural networks to improve trading systems based on technical analysis by means of the RSI financial indicator. Expert Systems with Applications 38, 9 (2011), 11489–11500.
    Google ScholarLocate open access versionFindings
  • Abhijit Sharang and Chetan Rao. 2015. Using machine learning for medium frequency derivative portfolio trading. Papers (2015).
    Google ScholarLocate open access versionFindings
  • Lawrence Takeuchi and Yu-Ying Albert Lee. 2013. Applying deep learning to enhance momentum trading strategies in stocks. In Technical Report. Stanford University.
    Google ScholarFindings
  • Richard H Thaler. 2005. Advances in behavioral finance. Vol. 2. Princeton University Press.
    Google ScholarFindings
  • 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
  • Daixin Wang, Peng Cui, and Wenwu Zhu. 2016. Structural Deep Network Embedding. In ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. 1225–1234.
    Google ScholarLocate open access versionFindings
  • Yingzi Zhu and Guofu Zhou. 2009. Technical analysis: An asset allocation perspective on the use of moving averages. Journal of Financial Economics 92, 3 (2009), 519–544.
    Google ScholarLocate open access versionFindings
Full Text
Your rating :
0

 

Tags
Comments