Categories
News

Feng Li will be presenting at the ISBA 2021 meeting

Categories
News

Our dlsa paper is accepted in the Journal of Computational and Graphical Statistics

Categories
News

New Paper: Forecasting reconciliation with a top-down alignment of independent level forecasts

Categories
News

New paper: Exploring the representativeness of the M5 competition data

Categories
News

New paper: Exploring the social influence of Kaggle virtual community on the M5 competition

Categories
News

The fuma paper is accepted in Journal of the Operational Research Society

Our fuma paper is accepted in the Journal of the Operational Research Society.

Xiaoqian Wang, Yanfei Kang, Fotios Petropoulos, Feng Li (2021). The uncertainty estimation of feature-based forecast combinations (in press), Journal of the Operational Research Society. 

Working paper | R package ]

Forecasting is an indispensable element of operational research (OR) and an important aid to planning. The accurate estimation of the forecast uncertainty facilitates several operations management activities, predominantly in supporting decisions in inventory and supply chain management and effectively setting safety stocks. In this paper, we introduce a feature-based framework, which links the relationship between time series features and the interval forecasting performance into providing reliable interval forecasts. We propose an optimal threshold ratio searching algorithm and a new weight determination mechanism for selecting an appropriate subset of models and assigning combination weights for each time series tailored to the observed features. We evaluate our approach using a large set of time series from the M4 competition. Our experiments show that our approach significantly outperforms a wide range of benchmark models, both in terms of point forecasts as well as prediction intervals.

Categories
News

New Paper: Forecast with Forecasts: Diversity Matters

Authors: Yanfei Kang, Wei Cao, Fotios Petropoulos, Feng Li

Abstract: Forecast combination has been widely applied in the last few decades to improve forecast accuracy. In recent years, the idea of using time series features to construct forecast combination model has flourished in the forecasting area. Although this idea has been proved to be beneficial in several forecast competitions such as the M3 and M4 competitions, it may not be practical in many situations. For example, the task of selecting appropriate features to build forecasting models can be a big challenge for many researchers. Even if there is one acceptable way to define the features, existing features are estimated based on the historical patterns, which are doomed to change in the future, or infeasible in the case of limited historical data. In this work, we suggest a change of focus from the historical data to the produced forecasts to extract features. We calculate the diversity of a pool of models based on the corresponding forecasts as a decisive feature and use meta-learning to construct diversity-based forecast combination models. A rich set of time series are used to evaluate the performance of the proposed method. Experimental results show that our diversity-based forecast combination framework not only simplifies the modelling process but also achieves superior forecasting performance.

Links: Working Paper

Categories
News

We are presenting at ISF2020 Invited Session

Our lab members will be presenting our work at the invited session of the 40th International Symposium on Forecasting virtually.

Session: Forecast Combination

Time: October 26, Monday, 17:00-18:00 GMT+8

Detailed Schedule: https://whova.com/embedded/session/iiofe_202006/1323449/

Speakers

  • Yanfei Kang (Speaker) Associate Professor, School of Economics and Management, Beihang University
  • Xiaoqian Wang (Speaker) PhD student, Beihang University
  • Xixi Li (Speaker) The University of Manchester.
  • Feng Li (Speaker) Assistant Professor at School of Statistics and Mathematics, Central University of Finance and Economics

Chair: Yanfei Kang

Forecast with forecasts: diversity matters

  • Yanfei Kang (Speaker) Associate Professor, School of Economics and Management, Beihang University

Forecast combination has been widely applied in the last few decades to improve forecast accuracy. In recent years, the idea of using time series features to construct forecast combination model has flourished in the forecasting area. Although this idea has been proved to be beneficial in several forecast competitions such as the M3 and M4 competitions, it may not be practical in many situations. For example, the task of selecting appropriate features to build forecasting models can be a big challenge for many researchers, and the interpretation may also be obscure so that it is hard to get valuable information from them. Hence, it is crucially important to improve the interpretability of forecast combination, making it feasible in practical applications. In this work, we treat the diversity of a pool of algorithms as an alternative to state-of-the-art time series features, and use meta-learning to construct diversity-based forecast combination models. A rich set of time series are used to evaluate the performance of the proposed method. Experimental results show that our diversity-based combination forecasting framework not only simplifies the modeling process but also achieves superior forecasting performance.

Distributed ARIMA Models for Ultra-long Time Series

Providing forecasts for ultra-long time series plays a vital role in various activities, such as investment decisions, industrial production arrangements, and farm management. This paper develops a novel distributed forecasting framework to tackle challenges associated with forecasting ultra-long time series by utilizing the industry-standard MapReduce framework. The proposed model combination approach facilitates distributed time series forecasting by combining the local estimators of ARIMA (AutoRegressive Integrated Moving Average) models delivered from worker nodes and minimizing a global loss function. In this way, instead of unrealistically assuming the data generating process (DGP) of an ultra-long time series stays invariant, we make assumptions only on the DGP of subseries spanning shorter time periods. We investigate the performance of the proposed distributed ARIMA models on an electricity demand dataset. Compared to ARIMA models, our approach results in significantly improved forecasting accuracy and computational efficiency both in point forecasts and prediction intervals, especially for longer forecast horizons. Moreover, we explore some potential factors that may affect the forecasting performance of our approach.

Improving forecasting with sub-seasonal time series patterns

Time series forecasting plays an increasingly important role in modern business decisions. In today’s data-rich environment, people often want to choose the optimal forecasting model for their data. However, identifying the optimal model often requires professional knowledge and experience, making accurate forecasting a challenging task. To mitigate the importance of model selection, we propose a simple and reliable algorithm and successfully improve forecasting performance. Specifically, we construct multiple time series with different sub-seasons from the original time series. These derived series highlight different sub-seasonal patterns of the original time series, making it possible for the forecasting methods to capture diverse patterns and components of the data. Subsequently, we make extrapolation forecasts for these multiple time series separately with classical statistical models (ETS or ARIMA). Finally, forecasts of these multiple time series are averaged together with equal weights. Whether in point or interval predictions, we evaluate our approach on the widely used competition datasets M1, M3, and M4 and it improves the forecasting performance in total horizon compared with the benchmarks. We also study which pattern of time series is more suitable for our method.

Feature-based Bayesian Forecasting Model Averaging

  • Feng Li (Speaker) Assistant Professor at School of Statistics and Mathematics, Central University of Finance and Economics

In this work, we propose a feature-based Bayesian forecasting model averaging framework (febama). Our Bayesian framework estimates weights of the feature-based forecasting combination via a Bayesian log predictive score, in which the optimal forecasting combination is connected and determined by time-series features from historical information. In particular, we utilize the prior knowledge of the coefficients of time-series features. We use an efficient Bayesian variable selection method to weight important features that may affect the forecasting combinations. To this end, our approach has better interpretability compared to other black-box forecasting combination schemes. Our framework is more computational efficient because the log predictive score and time-series features are calculated in the offline phase. We apply our framework to stock market data and M4 competition data. Based on our structure, a simple maximum-a-posteriori scheme outperforms the optimal prediction pools (Geweke and Amisano, 2011) or simple averaging, and Bayesian variable selection further enhanced the forecasting performance.

Categories
News

The Dejavu paper is accepted in the Journal of Business Research

Our Dejavu paper is accepted in the Journal of Business Research.

Yanfei Kang, Evangelos Spiliotis, Fotios Petropoulos, Nikolaos Athiniotis, Feng Li, Vassilios Assimakopoulo (2020). Déjà vu: A data-centric forecasting approach through time series cross-similarity, Journal of Business Research. (In Press)

[Working Paper | Software]

Accurate forecasts are vital for supporting the decisions of modern companies. Forecasters typically select the most appropriate statistical model for each time series. However, statistical models usually presume some data generation process while making strong assumptions about the errors. In this paper, we present a novel data-centric approach — “forecasting with similarity”, which tackles model uncertainty in a model-free manner. Existing similarity-based methods focus on identifying similar patterns within the series, i.e., “self-similarity”. In contrast, we propose searching for similar patterns from a reference set, i.e., “cross-similarity”. Instead of extrapolating, the future paths of the similar series are aggregated to obtain the forecasts of the target series. Building on the cross-learning concept, our approach allows the application of similarity-based forecasting on series with limited lengths. We evaluate the approach using a rich collection of real data and show that it yields competitive accuracy in both points forecasts and prediction intervals.

Categories
News

New Paper: Distributed ARIMA Models for Ultra-long Time Series