Improving the accuracy of global forecasting models using time series data augmentation

Bandara, Kasun; Hewamalage, Hansika; Liu, Yuanhao; Kang, Yanfei; Bergmeir, Christoph

doi:10.1016/j.patcog.2021.108148

Cited by 108 publications

(42 citation statements)

References 31 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Forecasting is sometimes complicated because the downstream supply chain stakeholders cannot share the information to support the forecast. Many industries such as food, retail, mining, rail, energy, tourism, and cloud computing need to generate more accurate forecasts to provide a better foundation for determining short-term, medium-term, and long-term corporate targets (Bandara, Hewamalage, Liu, Kang, & Bergmeir, 2021). One of the forecasting methods is the qualitative forecasting method which relies on personal judgment, intuition, and subjective evaluation (Chopra & Meindl, 2016).…”

Section: (Neisyafitri and Ongkunaruk)mentioning

confidence: 99%

The Use of Intervention Approach in Individual and Aggregate Forecasting Methods for Burger Patties: A Case in Indonesia

Neisyafitri

Ongkunaruk

2022

J AGRIB RURAL DEV RES

View full text Add to dashboard Cite

The Indonesian beef consumption increases sharply during Ramadan and made a difference between supply and demand. The research aimed to study the demand pattern of burger patties and determine a suitable forecasting method compared between quantitative and intervention forecasting methods. The actual demand was intervened by experts based on reasons such as supply shortage, holidays, promotion, and government projects. The daily sales of burger patties were collected for a year. Then, the data were divided into training and testing data. Later, time-series forecasting was performed by software. Then, the best forecasting method for daily data was selected between Individual forecasting and Top-Down forecasting. Similarly, for weekly data, the best forecasting method was compared between aggregate forecasting and Bottom-Up forecasting. Then, repeat the process for the intervened sales data. The result revealed that the mean absolute percentage error was improved after intervention by about 3.64%-58.83%. The combination of quantitative and qualitative approaches improved forecast accuracy. In addition, the aggregate level or weekly sales forecast had higher forecast accuracy than the disaggregated level. The Bottom-Up forecast performs better than the aggregate forecast. Hence, we recommended the company plans based on weekly data and implement Every Low Price to reduce the demand fluctuation.

show abstract

Section: (Neisyafitri and Ongkunaruk)mentioning

confidence: 99%

The Use of Intervention Approach in Individual and Aggregate Forecasting Methods for Burger Patties: A Case in Indonesia

Neisyafitri

Ongkunaruk

2022

J AGRIB RURAL DEV RES

View full text Add to dashboard Cite

show abstract

“…Data augmentation for time-series. Prior research on time-series data augmentation includes: (1) large-scale surveys exploring the impact of augmentation on various downstream modalities (Iwana and Uchida, 2021a,b; Wen et al, 2020); and (2) specific methods for particular modalities, including speech signals (Park et al, 2019(Park et al, , 2020, wearable device signals (Um et al, 2017), and time series forecasting (Bandara et al, 2021;Smyl and Kuber, 2016). There is relatively little work exploring how augmentation can impact performance for ECG-based prediction tasks, with prior studies mostly restricted to considering single tasks (Hatamian et al, 2020;Banerjee and Ghose, 2021).…”

Section: Related Workmentioning

confidence: 99%

Data Augmentation for Electrocardiograms

Raghu¹,

Shanmugam²,

Pomerantsev³

et al. 2022

Preprint

View full text Add to dashboard Cite

Neural network models have demonstrated impressive performance in predicting pathologies and outcomes from the 12-lead electrocardiogram (ECG). However, these models often need to be trained with large, labelled datasets, which are not available for many predictive tasks of interest. In this work, we perform an empirical study examining whether training time data augmentation methods can be used to improve performance on such datascarce ECG prediction problems. We investigate how data augmentation strategies impact model performance when detecting cardiac abnormalities from the ECG. Motivated by our finding that the effectiveness of existing augmentation strategies is highly task-dependent, we introduce a new method, TaskAug, which defines a flexible augmentation policy that is optimized on a per-task basis. We outline an efficient learning algorithm to do so that leverages recent work in nested optimization and implicit differentiation. In experiments, considering three datasets and eight predictive tasks, we find that TaskAug is competitive with or improves on prior work, and the learned policies shed light on what transformations are most effective for different tasks. We distill key insights from our experimental evaluation, generating a set of best practices for applying data augmentation to ECG prediction problems.Data and Code Availability We use three datasets: two are from Massachusetts General Hospital (MGH) and are not publicly available; the third is PTB-XL (Wagner et al., 2020), which is publicly available on the PhysioNet repository (Goldberger et al., 2000). Code implementing our method is available here: https://github.com/aniruddhraghu/ ecg_aug.

show abstract

“…Discarding rows of data with missing values will not affect most ML and DL models' training processes. However, some models require continuous sequences of data, such as RNNs [57]. In this case, a more in-depth analysis must be performed since time windows with missing values will hinder the performance of the trained model.…”

Section: Missing Values and Resamplingmentioning

confidence: 99%

Big Machinery Data Preprocessing Methodology for Data-Driven Models in Prognostics and Health Management

Cofre-Martel

Droguett

Modarres

2021

Sensors

View full text Add to dashboard Cite

Sensor monitoring networks and advances in big data analytics have guided the reliability engineering landscape to a new era of big machinery data. Low-cost sensors, along with the evolution of the internet of things and industry 4.0, have resulted in rich databases that can be analyzed through prognostics and health management (PHM) frameworks. Several data-driven models (DDMs) have been proposed and applied for diagnostics and prognostics purposes in complex systems. However, many of these models are developed using simulated or experimental data sets, and there is still a knowledge gap for applications in real operating systems. Furthermore, little attention has been given to the required data preprocessing steps compared to the training processes of these DDMs. Up to date, research works do not follow a formal and consistent data preprocessing guideline for PHM applications. This paper presents a comprehensive step-by-step pipeline for the preprocessing of monitoring data from complex systems aimed for DDMs. The importance of expert knowledge is discussed in the context of data selection and label generation. Two case studies are presented for validation, with the end goal of creating clean data sets with healthy and unhealthy labels that are then used to train machinery health state classifiers.

show abstract

Improving the accuracy of global forecasting models using time series data augmentation

Cited by 108 publications

References 31 publications

The Use of Intervention Approach in Individual and Aggregate Forecasting Methods for Burger Patties: A Case in Indonesia

The Use of Intervention Approach in Individual and Aggregate Forecasting Methods for Burger Patties: A Case in Indonesia

Data Augmentation for Electrocardiograms

Big Machinery Data Preprocessing Methodology for Data-Driven Models in Prognostics and Health Management

Contact Info

Product

Resources

About