BeatLex: Summarizing and Forecasting Time Series with Patterns

@inproceedings{Hooi2017BeatLexSA,
  title={BeatLex: Summarizing and Forecasting Time Series with Patterns},
  author={Bryan Hooi and Shenghua Liu and Asim Smailagic and Christos Faloutsos},
  booktitle={ECML/PKDD},
  year={2017}
}
Given time-series data such as electrocardiogram (ECG) readings, or motion capture data, how can we succintly summarize the data in a way that robustly identifies patterns that appear repeatedly? How can we then use such a summary to identify anomalies such as abnormal heartbeats, and also forecast future values of the time series? Our main idea is a vocabulary-based approach, which automatically learns a set of common patterns, or ‘beat patterns,’ which are used as building blocks to describe… 
BeatGAN: Anomalous Rhythm Detection using Adversarially Generated Time Series
TLDR
Experiments show that BeatGAN accurately and efficiently detects anomalous beats in ECG time series, and routes doctors’ attention to anomalous time ticks, achieving accuracy of nearly 0.95 AUC, and very fast inference (2.6 ms per beat).
Time Series Segmentation Based on Stationarity Analysis to Improve New Samples Prediction
TLDR
Two proposals to automatically segment subsequences in a time series were developed using change detection methods of data stream mining and the Long short-term memory predictive model reduced the relative prediction error from 1 to 0.67, compared to time series without segmentation.
Non-Linear Mining of Social Activities in Tensor Streams
TLDR
A streaming method that is designed to capture basic trends and seasonality in tensor streams and extract temporal and multi-dimensional relationships between such dynamics, and outperforms the state-of-the-art algorithms for time series forecasting in terms of forecasting accuracy and computational time.
StreamScope: Automatic Pattern Discovery over Data Streams
TLDR
This paper proposes a streaming algorithm, namely StreamScope, that is designed to find intuitive patterns efficiently from event streams evolving over time and can achieve great improvements in terms of computational time and memory space over its full batch method competitors.
Data Stream Analysis of Online Activities
  • Kouki Kawabata, Yasuko Matsubara, Yasushi Sakurai
  • Computer Science
    2018 IEEE International Conference on Internet of Things (iThings) and IEEE Green Computing and Communications (GreenCom) and IEEE Cyber, Physical and Social Computing (CPSCom) and IEEE Smart Data (SmartData)
  • 2018
TLDR
A streaming algorithm, namely STREAMSCOPE, that can find intuitive patterns efficiently from event streams evolving over time and achieve great improvements in terms of computational time and memory space over its full batch method competitors.
Anobeat: Anomaly Detection for Electrocardiography Beat Signals
TLDR
This paper proposes Anobeat, a semi-supervised approach, to perform the abnormal beat detection by facilitating adversarial regularized autoencoders constrained with multifeature and reconstruction error, which achieves ROC-AUC of 0.960 and 0.894 in the MIT-BIH intrapatient and inter-patient dataset respectively.
Multi-scale Anomaly Detection for Big Time Series of Industrial Sensors
TLDR
This work proposes a reconstruction-based anomaly detection method, MissGAN, iteratively learning to decode and encode naturally smooth time series in coarse segments, and can reconstruct a meaningful and robust time series, with the help of adversarial regularization and extra conditional states.
Stock Market Trend Prediction Using High-Order Information of Time Series
TLDR
A new method to simplify noisy-filled financial temporal series via sequence reconstruction by leveraging motifs (frequent patterns), and then utilize a convolutional neural network to capture spatial structure of time series is introduced.
Classical and Contemporary Approaches to Big Time Series Forecasting
TLDR
The objective of this tutorial is to provide a concise and intuitive overview of the most important methods and tools available for solving large-scale forecasting problems in three related fields: classical modeling of time series, scalable tensor methods, and deep learning for forecasting.
Stock Price Trend Prediction using MRCM-CNN
TLDR
A new method, MRCM-CNN, which combines time series reconstruction and deep neural network is introduced, which uses Spearman’s rank correlation as dependency measurement and performs segmentation and clustering simultaneously.
...
1
2
...

References

SHOWING 1-10 OF 30 REFERENCES
Finding semantics in time series
TLDR
This paper proposes to model time series data using a novel pattern-based hidden Markov model (pHMM), which aims at revealing a global picture of the system that generates the time seriesData, and proposes an iterative approach to refine pHMMs learned from the data.
AutoPlait: automatic mining of co-evolving time sequences
TLDR
The method has the following properties: effectiveness: it operates on large collections of time-series, and finds similar segment groups that agree with human intuition, and scalability: it is linear with the input size, and thus scales up very well.
MDL-based time series clustering
TLDR
This work shows that the problem definition for time series clustering from streams currently used is inherently flawed, and a new definition is necessary, and it is shown that the minimum description length framework offers an efficient, effective, and essentially parameter-free method for timeseries clustering.
HOT SAX: efficiently finding the most unusual time series subsequence
TLDR
The utility of discords with objective experiments on domains as diverse as Space Shuttle telemetry monitoring, medicine, surveillance, and industry, and the effectiveness of the discord discovery algorithm with more than one million experiments, on 82 different datasets from diverse domains are demonstrated.
Regime Shifts in Streams: Real-time Forecasting of Co-evolving Time Sequences
TLDR
REGIMECAST is designed as an adaptive non-linear dynamical system, which is inspired by the concept of "regime shifts" in natural dynamical systems and outperforms state-of-the-art competitors as regards accuracy and speed.
Clustering of time-series subsequences is meaningless: implications for previous and future research
TLDR
A novel method is introduced that is able to meaningfully cluster subsequences on some time-series datasets, based on the concept of time- series motifs, which invalidates the contribution of dozens of previously published papers.
Disk aware discord discovery: finding unusual time series in terabyte sized datasets
TLDR
This work shows how one particular definition of unusual time series, the time series discord, can be discovered with a disk aware algorithm, and uses the algorithm to provide further evidence of the effectiveness of the discord definition in areas as diverse as astronomy, web query mining, video surveillance, etc., and shows the efficiency of the method on datasets which are many orders of magnitude larger than anything else attempted.
DynaMMo: mining and summarization of coevolving sequences with missing values
TLDR
Given multiple time sequences with missing values, DynaMMo is proposed which summarizes, compresses, and finds latent variables, making the algorithm able to function even when there are missing values.
Forecasting Time Series With Complex Seasonal Patterns Using Exponential Smoothing
An innovations state space modeling framework is introduced for forecasting complex seasonal time series such as those with multiple seasonal periods, high-frequency seasonality, non-integer
An online algorithm for segmenting time series
TLDR
This paper undertake the first extensive review and empirical comparison of all proposed techniques for mining time-series data with fatal flaws and introduces a novel algorithm that is empirically show to be superior to all others in the literature.
...
1
2
3
...