Résultats de recherche pour « Time Series Forecasting »

Numerous deep learning architectures have been developed to accommodate the diversity of time-series datasets across different domains. In this article, we survey common encoder and decoder designs used in both one-step-ahead and multi-horizon time-series forecasting-describing how temporal information is incorporated into predictions by each model. Next, we highlight recent developments in hybrid deep learning models, which combine well-studied statistical models with neural network components to improve pure methods in either category. Lastly, we outline some ways in which deep learning can also facilitate decision support with time-series data. This article is part of the theme issue 'Machine learning for weather and climate modelling'.

Time series forecasting is one of the most widely used applications of data science. This chapter provides a comprehensive overview of time series analysis and forecasting. It starts by pointing out the distinction between standard supervised predictive models and time series forecasting models. It provides an introduction to the different time series forecasting methods, starting with time series decomposition, data-driven moving averages, exponential smoothing, discusses model-driven forecasts including regression, Autoregressive Integrated Moving Average methods, and machine learning-based methods using windowing techniques. The performance evaluation section provides a summary of key metrics to measure and compare the accuracy of the forecasting models. This chapter covers the implementation of forecasting techniques using RapidMiner and R.

Recently, there has been a surge of Transformer-based solutions for the long-term time series forecasting (LTSF) task. Despite the growing performance over the past few years, we question the validity of this line of research in this work. Specifically, Transformers is arguably the most successful solution to extract the semantic correlations among the elements in a long sequence. However, in time series modeling, we are to extract the temporal relations in an ordered set of continuous points. While employing positional encoding and using tokens to embed sub-series in Transformers facilitate preserving some ordering information, the nature of the permutation-invariant self-attention mechanism inevitably results in temporal information loss. To validate our claim, we introduce a set of embarrassingly simple one-layer linear models named LTSF-Linear for comparison. Experimental results on nine real-life datasets show that LTSF-Linear surprisingly outperforms existing sophisticated Transformer-based LTSF models in all cases, and often by a large margin. Moreover, we conduct comprehensive empirical studies to explore the impacts of various design elements of LTSF models on their temporal relation extraction capability. We hope this surprising finding opens up new research directions for the LTSF task. We also advocate revisiting the validity of Transformer-based solutions for other time series analysis tasks (e.g., anomaly detection) in the future.

Abstract In this paper, we survey the most recent advances in supervised machine learning (ML) and high‐dimensional models for time‐series forecasting. We consider both linear and nonlinear alternatives. Among the linear methods, we pay special attention to penalized regressions and ensemble of models. The nonlinear methods considered in the paper include shallow and deep neural networks, in their feedforward and recurrent versions, and tree‐based methods, such as random forests and boosted trees. We also consider ensemble and hybrid models by combining ingredients from different alternatives. Tests for superior predictive ability are briefly reviewed. Finally, we discuss application of ML in economics and finance and provide an illustration with high‐frequency financial data.

Abstract Time series forecasting (TSF) has long been a crucial task in both industry and daily life. Most classical statistical models may have certain limitations when applied to practical scenarios in fields such as energy, healthcare, traffic, meteorology, and economics, especially when high accuracy is required. With the continuous development of deep learning, numerous new models have emerged in the field of time series forecasting in recent years. However, existing surveys have not provided a unified summary of the wide range of model architectures in this field, nor have they given detailed summaries of works in feature extraction and datasets. To address this gap, in this review, we comprehensively study the previous works and summarize the general paradigms of Deep Time Series Forecasting (DTSF) in terms of model architectures. Besides, we take an innovative approach by focusing on the composition of time series and systematically explain important feature extraction methods. Additionally, we provide an overall compilation of datasets from various domains in existing works. Finally, we systematically emphasize the significant challenges faced and future research directions in this field.

The recent boom of linear forecasting models questions the ongoing passion for architectural modifications of Transformer-based forecasters. These forecasters leverage Transformers to model the global dependencies over temporal tokens of time series, with each token formed by multiple variates of the same timestamp. However, Transformers are challenged in forecasting series with larger lookback windows due to performance degradation and computation explosion. Besides, the embedding for each temporal token fuses multiple variates that represent potential delayed events and distinct physical measurements, which may fail in learning variate-centric representations and result in meaningless attention maps. In this work, we reflect on the competent duties of Transformer components and repurpose the Transformer architecture without any modification to the basic components. We propose iTransformer that simply applies the attention and feed-forward network on the inverted dimensions. Specifically, the time points of individual series are embedded into variate tokens which are utilized by the attention mechanism to capture multivariate correlations; meanwhile, the feed-forward network is applied for each variate token to learn nonlinear representations. The iTransformer model achieves state-of-the-art on challenging real-world datasets, which further empowers the Transformer family with promoted performance, generalization ability across different variates, and better utilization of arbitrary lookback windows, making it a nice alternative as the fundamental backbone of time series forecasting. Code is available at this repository: https://github.com/thuml/iTransformer.

In this paper, we study the usage of machine-learning models for sales predictive analytics. The main goal of this paper is to consider main approaches and case studies of using machine learning for sales forecasting. The effect of machine-learning generalization has been considered. This effect can be used to make sales predictions when there is a small amount of historical data for specific sales time series in the case when a new product or store is launched. A stacking approach for building regression ensemble of single models has been studied. The results show that using stacking techniques, we can improve the performance of predictive models for sales time series forecasting.

For robust detection performance, traditional control chart monitoring for biosurveillance is based on input data free of trends, day-of-week effects, and other systematic behaviour. Time series forecasting methods may be used to remove this behaviour by subtracting forecasts from observations to form residuals for algorithmic input. We describe three forecast methods and compare their predictive accuracy on each of 16 authentic syndromic data streams. The methods are (1) a non-adaptive regression model using a long historical baseline, (2) an adaptive regression model with a shorter, sliding baseline, and (3) the Holt-Winters method for generalized exponential smoothing. Criteria for comparing the forecasts were the root-mean-square error, the median absolute per cent error (MedAPE), and the median absolute deviation. The median-based criteria showed best overall performance for the Holt-Winters method. The MedAPE measures over the 16 test series averaged 16.5, 11.6, and 9.7 for the non-adaptive regression, adaptive regression, and Holt-Winters methods, respectively. The non-adaptive regression forecasts were degraded by changes in the data behaviour in the fixed baseline period used to compute model coefficients. The mean-based criterion was less conclusive because of the effects of poor forecasts on a small number of calendar holidays. The Holt-Winters method was also most effective at removing serial autocorrelation, with most 1-day-lag autocorrelation coefficients below 0.15. The forecast methods were compared without tuning them to the behaviour of individual series. We achieved improved predictions with such tuning of the Holt-Winters method, but practical use of such improvements for routine surveillance will require reliable data classification methods.

Automatic forecasts of large numbers of univariate time series are often needed in business and other contexts. We describe two automatic forecasting algorithms that have been implemented in the forecast package for R. The first is based on innovations state space models that underly exponential smoothing methods. The second is a step-wise algorithm for forecasting with ARIMA models. The algorithms are applicable to both seasonal and non-seasonal data, and are compared and illustrated using four real time series. We also briefly describe some of the other functionality available in the forecast package.