Similarity-Based Data-Fusion Schemes for Missing Data Imputation in Univariate Time Series Data

In cross-sectional time-series data with a dichotomous dependent variable, failing to account for duration dependence when it exists can lead to faulty inferences. A common solution is to include duration dummies, polynomials, or splines to proxy for duration dependence. Because creating these is not easy for the common practitioner, I introduce a new command, mkduration, that is a straightforward way to generate a duration variable for binary cross-sectional time-series data in Stata. mkduration can handle various forms of missing data and allows the duration variable to easily be turned into common parametric and nonparametric approximations.

Download Full-text

Validation of synthetic daily Landsat NDVI time series data generated by the improved spatial and temporal data fusion approach

Information Fusion ◽

10.1016/j.inffus.2017.06.005 ◽

2018 ◽

Vol 40 ◽

pp. 34-44 ◽

Cited By ~ 14

Author(s):

Mingquan Wu ◽

Wenjiang Huang ◽

Zheng Niu ◽

Changyao Wang ◽

Wang Li ◽

...

Keyword(s):

Time Series ◽

Data Fusion ◽

Time Series Data ◽

Series Data ◽

Temporal Data ◽

Ndvi Time Series ◽

Fusion Approach

Download Full-text

An Empirical Mode-Spatial Model for Environmental Data Imputation

Hydrology ◽

10.3390/hydrology5040063 ◽

2018 ◽

Vol 5 (4) ◽

pp. 63 ◽

Cited By ~ 1

Author(s):

Benjamin Nelsen ◽

D. Williams ◽

Gustavious Williams ◽

Candace Berrett

Keyword(s):

Time Series ◽

Spatial Data ◽

Missing Values ◽

Time Series Data ◽

Environmental Data ◽

Series Data ◽

Data Imputation ◽

Accurate Data ◽

Target Station ◽

Periodic Components

Complete and accurate data are necessary for analyzing and understanding trends in time-series datasets; however, many of the available time-series datasets have gaps that affect the analysis, especially in the earth sciences. As most available data have missing values, researchers use various interpolation methods or ad hoc approaches to data imputation. Since the analysis based on inaccurate data can lead to inaccurate conclusions, more accurate data imputation methods can provide accurate analysis. We present a spatial-temporal data imputation method using Empirical Mode Decomposition (EMD) based on spatial correlations. We call this method EMD-spatial data imputation or EMD-SDI. Though this method is applicable to other time-series data sets, here we demonstrate the method using temperature data. The EMD algorithm decomposes data into periodic components called intrinsic mode functions (IMF) and exactly reconstructs the original signal by summing these IMFs. EMD-SDI initially decomposes the data from the target station and other stations in the region into IMFs. EMD-SDI evaluates each IMF from the target station in turn and selects the IMF from other stations in the region with periodic behavior most correlated to target IMF. EMD-SDI then replaces a section of missing data in the target station IMF with the section from the most closely correlated IMF from the regional stations. We found that EMD-SDI selects the IMFs used for reconstruction from different stations throughout the region, not necessarily the station closest in the geographic sense. EMD-SDI accurately filled data gaps from 3 months to 5 years in length in our tests and favorably compares to a simple temporal method. EMD-SDI leverages regional correlation and the fact that different stations can be subject to different periodic behaviors. In addition to data imputation, the EMD-SDI method provides IMFs that can be used to better understand regional correlations and processes.

Download Full-text

Missing Data Imputation and Meta-analysis on Correlation of Spatio-Temporal Weather Series Data

10.1109/iemcon53756.2021.9623154 ◽

2021 ◽

Author(s):

Alkiviadis Kyrtsoglou ◽

Dimara Asimina ◽

Dimitrios Triantafyllidis ◽

Stelios Krinidis ◽

Konstantinos Kitsikoudis ◽

...

Keyword(s):

Missing Data ◽

Meta Analysis ◽

Series Data ◽

Data Imputation ◽

Missing Data Imputation ◽

Spatio Temporal

Download Full-text

Exploratory Time Series Data Mining by Genetic Clustering

Mathematical Methods for Knowledge Discovery and Data Mining ◽

10.4018/978-1-59904-528-3.ch010 ◽

2011 ◽

pp. 157-178

Author(s):

T. Warren Liao

Keyword(s):

Data Mining ◽

Time Series ◽

Time Series Data ◽

Distance Measures ◽

Series Data ◽

Synthetic Control ◽

Data Set ◽

Univariate Time Series ◽

Genetic Clustering ◽

Data Objects

In this chapter, we present genetic algorithm (GA) based methods developed for clustering univariate time series with equal or unequal length as an exploratory step of data mining. These methods basically implement the k-medoids algorithm. Each chromosome encodes in binary the data objects serving as the k-medoids. To compare their performance, both fixed-parameter and adaptive GAs were used. We first employed the synthetic control chart data set to investigate the performance of three fitness functions, two distance measures, and other GA parameters such as population size, crossover rate, and mutation rate. Two more sets of time series with or without known number of clusters were also experimented: one is the cylinder-bell-funnel data and the other is the novel battle simulation data. The clustering results are presented and discussed.

Download Full-text

A Metric Learning-Based Univariate Time Series Classification Method

Information ◽

10.3390/info11060288 ◽

2020 ◽

Vol 11 (6) ◽

pp. 288

Author(s):

Kuiyong Song ◽

Nianbin Wang ◽

Hongbin Wang

Keyword(s):

Time Series ◽

Time Series Data ◽

Multivariate Time Series ◽

Metric Learning ◽

Classification Method ◽

Series Data ◽

Classification Error ◽

Time Series Classification ◽

Classification Error Rate ◽

Univariate Time Series

High-dimensional time series classification is a serious problem. A similarity measure based on distance is one of the methods for time series classification. This paper proposes a metric learning-based univariate time series classification method (ML-UTSC), which uses a Mahalanobis matrix on metric learning to calculate the local distance between multivariate time series and combines Dynamic Time Warping(DTW) and the nearest neighbor classification to achieve the final classification. In this method, the features of the univariate time series are presented as multivariate time series data with a mean value, variance, and slope. Next, a three-dimensional Mahalanobis matrix is obtained based on metric learning in the data. The time series is divided into segments of equal intervals to enable the Mahalanobis matrix to more accurately describe the features of the time series data. Compared with the most effective measurement method, the related experimental results show that our proposed algorithm has a lower classification error rate in most of the test datasets.

Download Full-text

A dual-head attention model for time series data imputation

Computers and Electronics in Agriculture ◽

10.1016/j.compag.2021.106377 ◽

2021 ◽

Vol 189 ◽

pp. 106377

Author(s):

Yifan Zhang ◽

Peter J. Thorburn

Keyword(s):

Time Series ◽

Time Series Data ◽

Series Data ◽

Data Imputation ◽

Attention Model

Download Full-text

Time-Series Causality with Missing Data

Journal of Computational Vision and Imaging Systems ◽

10.15353/jcvis.v6i1.3552 ◽

2021 ◽

Vol 6 (1) ◽

pp. 1-4

Author(s):

Bo Yuan Chang ◽

Mohamed A. Naiel ◽

Steven Wardell ◽

Stan Kleinikkink ◽

John S. Zelek

Keyword(s):

Time Series ◽

Missing Data ◽

Missing Values ◽

Time Series Data ◽

Multivariate Time Series ◽

Gaussian Process Regression ◽

Series Data ◽

Causal Relationships ◽

Sampled Data ◽

The Past

Over the past years, researchers have proposed various methods to discover causal relationships among time-series data as well as algorithms to fill in missing entries in time-series data. Little to no work has been done in combining the two strategies for the purpose of learning causal relationships using unevenly sampled multivariate time-series data. In this paper, we examine how the causal parameters learnt from unevenly sampled data (with missing entries) deviates from the parameters learnt using the evenly sampled data (without missing entries). However, to obtain the causal relationship from a given time-series requires evenly sampled data, which suggests filling the missing data values before obtaining the causal parameters. Therefore, the proposed method is based on applying a Gaussian Process Regression (GPR) model for missing data recovery, followed by several pairwise Granger causality equations in Vector Autoregssive form to fit the recovered data and obtain the causal parameters. Experimental results show that the causal parameters generated by using GPR data filling offers much lower RMSE than the dummy model (fill with last seen entry) under all missing values percentage, suggesting that GPR data filling can better preserve the causal relationships when compared with dummy data filling, thus should be considered when dealing with unevenly sampled time-series causality learning.

Download Full-text