arXiv subjects

Rob J Hyndman

Publications and source records attributed to Rob J Hyndman.

23 records · Page 2Linked to original sources

arXiv2020

GRATIS: GeneRAting TIme Series with diverse and controllable characteristics

The explosion of time series data in recent years has brought a flourish of new time series analysis methods, for forecasting, clustering, classification and other tasks. The evaluation of these new methods requires either collecting or simulating a diverse set of time series benchmarking data to enable reliable comparisons against alternative approaches. We propose GeneRAting TIme Series with diverse and controllable characteristics, named GRATIS, with the use of mixture autoregressive (MAR) models. We simulate sets of time series using MAR models and investigate the diversity and coverage of the generated time series in a time series feature space. By tuning the parameters of the MAR models, GRATIS is also able to efficiently generate new time series with controllable features. In general, as a costless surrogate to the traditional data collection approach, GRATIS can be used as an evaluation tool for tasks such as time series forecasting and classification. We illustrate the usefulness of our time series generation process through a time series forecasting application.

stat.ML↗

arXiv2019

Machine learning applications in time series hierarchical forecasting

Hierarchical forecasting (HF) is needed in many situations in the supply chain (SC) because managers often need different levels of forecasts at different levels of SC to make a decision. Top-Down (TD), Bottom-Up (BU) and Optimal Combination (COM) are common HF models. These approaches are static and often ignore the dynamics of the series while disaggregating them. Consequently, they may fail to perform well if the investigated group of time series are subject to large changes such as during the periods of promotional sales. We address the HF problem of predicting real-world sales time series that are highly impacted by promotion. We use three machine learning (ML) models to capture sales variations over time. Artificial neural networks (ANN), extreme gradient boosting (XGboost), and support vector regression (SVR) algorithms are used to estimate the proportions of lower-level time series from the upper level. We perform an in-depth analysis of 61 groups of time series with different volatilities and show that ML models are competitive and outperform some well-established models in the literature.

cs.LG↗

arXiv2019

A new tidy data structure to support exploration and modeling of temporal data

Mining temporal data for information is often inhibited by a multitude of formats: irregular or multiple time intervals, point events that need aggregating, multiple observational units or repeated measurements on multiple individuals, and heterogeneous data types. On the other hand, the software supporting time series modeling and forecasting, makes strict assumptions on the data to be provided, typically requiring a matrix of numeric data with implicit time indexes. Going from raw data to model-ready data is painful. This work presents a cohesive and conceptual framework for organizing and manipulating temporal data, which in turn flows into visualization, modeling and forecasting routines. Tidy data principles are extended to temporal data by: (1) mapping the semantics of a dataset into its physical layout; (2) including an explicitly declared index variable representing time; (3) incorporating a "key" comprising single or multiple variables to uniquely identify units over time. This tidy data representation most naturally supports thinking of operations on the data as building blocks, forming part of a "data pipeline" in time-based contexts. A sound data pipeline facilitates a fluent workflow for analyzing temporal data. The infrastructure of tidy temporal data has been implemented in the R package "tsibble".

stat.AP↗

arXiv2018

Calendar-based graphics for visualizing people's daily schedules

Calendars are broadly used in society to display temporal information, and events. This paper describes a new R package with functionality to organize and display temporal data, collected on sub-daily resolution, into a calendar layout. The function `frame_calendar` uses linear algebra on the date variable to restructure data into a format lending itself to calendar layouts. The user can apply the grammar of graphics to create plots inside each calendar cell, and thus the displays synchronize neatly with ggplot2 graphics. The motivating application is studying pedestrian behavior in Melbourne, Australia, based on counts which are captured at hourly intervals by sensors scattered around the city. Faceting by the usual features such as day and month, was insufficient to examine the behavior. Making displays on a monthly calendar format helps to understand pedestrian patterns relative to events such as work days, weekends, holidays, and special events. The layout algorithm has several format options and variations. It is implemented in the R package sugrrants.

stat.AP↗

arXiv2016

Grouped functional time series forecasting: An application to age-specific mortality rates

Age-specific mortality rates are often disaggregated by different attributes, such as sex, state and ethnicity. Forecasting age-specific mortality rates at the national and sub-national levels plays an important role in developing social policy. However, independent forecasts at the sub-national levels may not add up to the forecasts at the national level. To address this issue, we consider reconciling forecasts of age-specific mortality rates, extending the methods of Hyndman et al. (2011) to functional time series, where age is considered as a continuum. The grouped functional time series methods are used to produce point forecasts of mortality rates that are aggregated appropriately across different disaggregation factors. For evaluating forecast uncertainty, we propose a bootstrap method for reconciling interval forecasts. Using the regional age-specific mortality rates in Japan, obtained from the Japanese Mortality Database, we investigate the one- to ten-step-ahead point and interval forecast accuracies between the independent and grouped functional time series forecasting methods. The proposed methods are shown to be useful for reconciling forecasts of age-specific mortality rates at the national and sub-national levels. They also enjoy improved forecast accuracy averaged over different disaggregation factors. Supplemental materials for the article are available online.

stat.AP↗