Abstract:Multivariate time series data are becoming increasingly common in numerous real world applications, e.g., power plant monitoring, health care, wearable devices, automobile, etc. As a result, multivariate time series retrieval, i.e., given the current multivariate time series segment, how to obtain its relevant time series segments in the historical data (or in the database), attracts significant amount of interest in many fields. Building such a system, however, is challenging since it requires a compact representation of the raw time series which can explicitly encode the temporal dynamics as well as the correlations (interactions) between different pairs of time series (sensors). Furthermore, it requires query efficiency and expects a returned ranking list with high precision on the top. Despite the fact that various approaches have been developed, few of them can jointly resolve these two challenges. To cope with this issue, in this paper we propose a Deep r-th root of Rank Supervised Joint Binary Embedding (Deep r-RSJBE) to perform multivariate time series retrieval. Given a raw multivariate time series segment, we employ Long Short-Term Memory (LSTM) units to encode the temporal dynamics and utilize Convolutional Neural Networks (CNNs) to encode the correlations (interactions) between different pairs of time series (sensors). Subsequently, a joint binary embedding is pursued to incorporate both the temporal dynamics and the correlations. Finally, we develop a novel r-th root ranking loss to optimize the precision at the top of a Hamming distance ranking list. Thoroughly empirical studies based upon three publicly available time series datasets demonstrate the effectiveness and the efficiency of Deep r-RSJBE.

REGER: Reordering Time Series Data for Regression Encoding

Time Series Data Encoding for Efficient Storage

A novel piecewise linear segmentation for time series

Time series data encoding in Apache IoTDB: comparative analysis and recommendation

Time Series Compression based on Reinforcement Learning

TRACE: Real-time Compression of Streaming Trajectories in Road Networks

RegHD: Robust and Efficient Regression in Hyper-Dimensional Learning System

Frequency Domain Data Encoding in Apache IoTDB.

Multimodal Data Encoding and Compression in Apache IoTDB

SHRINK: Data Compression by Semantic Extraction and Residuals Encoding

In-Network Time-Series Data Compression for Electric Internet of Things

MOST: Model-Based Compression with Outlier Storage for Time Series Data

Retrieval Based Time Series Forecasting

Backward-Sort for Time Series in Apache IoTDB.

Time Series Compression Based on Adaptive Piecewise Recurrent Autoencoder

Class-Based Delta-Encoding for High-Speed Train Data Stream

Segmentation of Time Series Based on Kinetic Characteristics for Storage Consumption Prediction

Elf: Erasing-based Lossless Floating-Point Compression.

Deep Dict: Deep Learning-based Lossy Time Series Compressor for IoT Data

TVStore: Automatically Bounding Time Series Storage via Time-Varying Compression

Deep R-Th Root of Rank Supervised Joint Binary Embedding for Multivariate Time Series Retrieval