• Corpus ID: 14163167

Pattern discovery from stock time series using self-organizing maps

@inproceedings{Fu2016PatternDF,
  title={Pattern discovery from stock time series using self-organizing maps},
  author={Tak-Chung Fu and Fu-lai Chung and Robert Wing Pong Luk and Ng},
  year={2016}
}
† This work was supported by the RGC CERG project PolyU 5065/98E and the Departmental Grant H-ZJ84 ‡ Corresponding author ABSTRACT Pattern discovery from time series is of fundamental importance. Particularly when the domain expert derived patterns do not exist or are not complete, an algorithm to discover specific patterns or shapes automatically from the time series data is necessary. Such an algorithm is noteworthy in that it does not assume prior knowledge of the number of interesting… 

Figures and Tables from this paper

Feature Clustering with Self-organizing Maps and an Application to Financial Time-series for Portfolio Selection
TLDR
Results show that feature clustering with the SOM presents itself as a viable method to cluster time-series, and aims to develop a tool to assist the investor in finding balanced portoflios.
Effective Clustering of Time-Series Data Using FCM
TLDR
A two-level fuzzy clustering strategy is employed in order to achieve the objective of addressing the problem of time series clustering through conventional approach with benefits presented by implementing a real application: Credit card Transactions Clustering.
(Not) Finding Rules in Time Series: A Surprising Result with Implications for Previous and Future Research
Time series data is perhaps the most frequently encountered type of data examined by the data mining community. Clustering is perhaps the most frequently used data mining algorithm, being useful in
Automatically Recognizing Stock Patterns Using RPCL Neural Networks
TLDR
An improved version of the rival penalized competitive learning (RPCL) is introduced, and a comparative study between the clustering performances of the improved RPCL and the SOM is conducted, showing that a better clustering performance can be achieved by the former.
Pattern Discovery in Time Series A survey
TLDR
This work ordered time series in terms of their theoretical maximum predictability and showed that the accuracy of various forecasting algorithms improves with the increase of the value of predictability with the latter being a theoretical limit of accuracy.
Shape-based time series similarity measure and pattern discovery algorithm
TLDR
A similarity measure based on shape, Sh measure, is originally proposed, and the properties of this similarity and corresponding proofs are given, and a time series shape pattern discovery algorithm based on Sh measure is put forward.
Recent Techniques of Clustering of Time Series Data: A Survey
TLDR
This paper has shown the survey and summarization of previous work that investigated the clustering of time series in various application domains ranging from science, engineering, business, finance, economic, health care, and government.
Intelligent Trading System: Multidimensional financial time series clustering based on self-organizing map
TLDR
This paper proposes a multidimensional time series clustering model based on graph attention autoencoder and mask self-organizing map (Mask-SOM), based on which it realizes multi-step prediction of financial derivatives prices and intelligent trading system construction.
...
...

References

SHOWING 1-10 OF 26 REFERENCES
Evolutionary segmentation of financial time series into subsequences
TLDR
This paper proposes an evolutionary segmentation algorithm and encouraging experimental results in segmenting the Hong Kong Hang Seng Index using 22 technical analysis patterns are reported.
Non-stationary time-series prediction using fuzzy clustering
  • A. Geva
  • Computer Science
    18th International Conference of the North American Fuzzy Information Processing Society - NAFIPS (Cat. No.99TH8397)
  • 1999
TLDR
One of the advantages of this new algorithm is its adaptive hierarchical selection of the number of clusters, which can overcome the general non-stationary nature of real-life time-series (biomedical, physical, economical, etc.).
Rule Discovery from Time Series
We consider the problem of finding rules relating patterns in a time series to other patterns in that series, or patterns in one series to patterns in another series. A simple example is a rule such
DATA MINING OF MULTIPLE NONSTATIONARY TIME SERIES
TLDR
A data mining method that embeds multiple time series into a phase space and chooses an optimal local model for short-term forecasting produces better temporal patterns than single time series embedding.
Neural Networks for Pattern Recognition
Non-stationary signal analysis using temporal clustering
  • S. PolickerA. Geva
  • Computer Science
    Neural Networks for Signal Processing VIII. Proceedings of the 1998 IEEE Signal Processing Society Workshop (Cat. No.98TH8378)
  • 1998
TLDR
A model of nonstationary time series generated by switching between a finite number of random processes is presented and temporal clustering is applied to estimate the model's parameters and a drift between disjoint states is analyzed.
Component analysis in financial time series
  • R. H. LeschY. CailléD. Lowe
  • Economics
    Proceedings of the IEEE/IAFE 1999 Conference on Computational Intelligence for Financial Engineering (CIFEr) (IEEE Cat. No.99TH8408)
  • 1999
TLDR
The main objective is to find out if principal component analysis and independent component analysis are able to perform feature extraction, signal-noise-decomposition and dimensionality reduction, since that would enable a further inside look into the behaviour and mechanics of financial markets.
A new approach to transforming time series into symbolic sequences
  • Kai Ou-YangWenyan JiaPin ZhouXin Meng
  • Computer Science
    Proceedings of the First Joint BMES/EMBS Conference. 1999 IEEE Engineering in Medicine and Biology 21st Annual Conference and the 1999 Annual Fall Meeting of the Biomedical Engineering Society (Cat. N
  • 1999
TLDR
This work proposes a new approach to transform time series into symbolic sequences, which can maintain most of the information included in time series and gives an example of its application.
Fast Learning in Networks of Locally-Tuned Processing Units
We propose a network architecture which uses a single internal layer of locally-tuned processing units to learn both classification tasks and real-valued function approximations (Moody and Darken
Discriminative Nonlinear Dimensionality Reduction for Improved Classification
  • M. Dolson
  • Computer Science
    Int. J. Neural Syst.
  • 1994
TLDR
It is shown how recent advances in nonlinear dimensionality reduction can be incorporated into the indirect approach, and how the two approaches can be integrated in a novel MLP framework and it is shown that applying these ideas to the classification of temporal trajectories can substantially improve performance on simple tasks.
...
...