An approach to handle concept drift in financial time series based on Extreme Learning Machines and explicit Drift Detection

Cavalcante, Rodolfo C.; Oliveira, Adriano L. I.

doi:10.1109/ijcnn.2015.7280721

Cited by 29 publications

(28 citation statements)

References 24 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Such concept drift occurs in financial time series data [19]; one approach to addressing this is to recalibrate the models on a pre-determined basis (i.e. implicit); the other approach is to have a trigger (i.e.…”

Section: ) Stock Market States and Financial Forecastingmentioning

confidence: 99%

“…Among many other things this change over time can be driven by economic changes. Approaches to address its non-stationary nature have involved using the financial timeseries data of the company in question and apply clustering techniques to identify the point at which to recalibrate the forecasting model [19][20] to accommodate for the changes. However, these approaches tend to rely on the data of the company that is being forecast and not necessarily try to link it to the overall economic movements or moods of the stock market.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Stock Price Forecasting Incorporating Market State

Beyaz¹,

Tekiner²,

Zeng³

et al. 2018

2018 IEEE 20th International Conference on High Performance Computing and Communications; IEEE 16th International Conference On

View full text Add to dashboard Cite

This paper evaluates the impact of identifying and accounting explicitly for the states of the stock market when forecasting stock prices using machine learning. Clustering methods are applied to market mood indicators (e.g. VIX) to identify various states of the stock market and forecasting models for the selected companies are developed for each market state. Tests using 85 companies from the S&P 500 indicate that, for a forecasting horizon of 126 days, accounting for the market mood improves forecasting accuracy (lower RMSE) in 47% of cases. This improvement from accounting explicitly for market states is also seen when only Technical indicators or Fundamental indicators are inputs to the forecasting models; the benefit decreases when these indicators are combined.

show abstract

Section: ) Stock Market States and Financial Forecastingmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Stock Price Forecasting Incorporating Market State

Beyaz¹,

Tekiner²,

Zeng³

et al. 2018

2018 IEEE 20th International Conference on High Performance Computing and Communications; IEEE 16th International Conference On

View full text Add to dashboard Cite

show abstract

“…In a recent work [5], we investigated the use of online explicit drift detection in time series by means of monitoring the error of a regression model. In that work, we implemented the Drift Detection Mechanism (DDM) [11] and the Exponentially Weighted Moving Average for Concept Drift Detection (ECDD) [13] in combination with extreme learning machine (ELM) algorithm to build an adaptive prediction method.…”

Section: Related Workmentioning

confidence: 99%

“…In order to do so, we compare FEDD with the ELM ECDD method proposed in [5], since it is based on prediction error and uses the same drift test as FEDD. The second objective is to verify the drift identification accuracy of FEDD in comparison with other error-based drift detection methods.…”

Section: Computational Experimentsmentioning

confidence: 99%

See 1 more Smart Citation

FEDD: Feature Extraction for Explicit Concept Drift Detection in time series

Cavalcante

Minku

Oliveira

2016

2016 International Joint Conference on Neural Networks (IJCNN)

Self Cite

View full text Add to dashboard Cite

Abstract-A time series is a sequence of observations collected over fixed sampling intervals. Several real-world dynamic processes can be modeled as a time series, such as stock price movements, exchange rates, temperatures, among others. As a special kind of data stream, a time series may present concept drift, which affects negatively time series analysis and forecasting. Explicit drift detection methods based on monitoring the time series features may provide a better understanding of how concepts evolve over time than methods based on monitoring the forecasting error of a base predictor. In this paper, we propose an online explicit drift detection method that identifies concept drifts in time series by monitoring time series features, called Feature Extraction for Explicit Concept Drift Detection (FEDD). Computational experiments showed that FEDD performed better than error-based approaches in several linear and nonlinear artificial time series with abrupt and gradual concept drifts.

show abstract