Long-term Time Series Forecasting Research Articles

Handling Massive Proportion of Missing Labels in Multivariate Long-Term Time Series Forecasting

Training Deep Learning (DL) models with missing labels is a challenge in diverse engineering applications. Missing value imputation methods have been proposed to try to address this problem, but their performance is affected with Massive Proportion of Missing Labels (MPML). This paper presents a approach for handling MPML in Multivariate Long-Term Time Series Forecasting. It is an two-step process where interpolation (using Gaussian Processes Regression (GPR) and domain knowledge from experts) and prediction model are separated to enable the integration of prior domain knowledge. First, a set of samples of the possible interpolation of the missing outputs are generated by the GPR based on the domain knowledge. Second, the observed input sensor data and interpolated labels from GPR are used to train the prediction model. We evaluated our approach with the development of a soft-sensor with one real datasets to forecast the biomass during recombinant adeno-associated virus (rAAV) production in bioreactors. Our experimental results demonstrate the potential of the approach through quantitative evaluation of the generated forecasts in a case that would be extremely difficult to train a DL model due to MPML.

A deep granular network with adaptive unequal-length granulation strategy for long-term time series forecasting and its industrial applications

An accurate long-term forecasting for some time series in industrial production is substantially significant for improving the economic efficiency of industry enterprise. In this study, a granular computing (GrC)-based deep learning framework is proposed for long-term time series forecasting, which consists of two stages i.e., the adaptive data granulation and the GrC-based deep model construction. In the first stage, for automatically generating the basic information granules with unequal time span adaptively from data, a stacked sparse auto-encoders granulation network is designed, in which the starting and ending points of a granule are adaptively determined by setting a single neuron in the last hidden layer after multi-layer feature extraction. Then, the second stage sees a definition of a partially overlapping sub-block basis (POSB) matrix to extract the features of these granules, based on which a deep sparse coding feature decomposition-based long-term forecasting model is constructed to transform the unequal-length granules into a product of a POSB matrix and a coefficient matrix layer by layer to serve for forecasting. To verify the effectiveness of the proposed method, two synthetic datasets, two real-world datasets and two practical industrial datasets are employed. The experimental results demonstrate that the proposed method outperforms other data-driven ones on long-term time series forecasting, particularly in an industrial case.

Adaptive Temporal-Frequency Network for Time-Series Forecasting

A novel adaptive temporal-frequency network (ATFN), which is an end-to-end hybrid model incorporating deep learning networks and frequency patterns, is proposed for mid- and long-term time series forecasting. Within the framework of the ATFN, an augmented sequence to sequence model is used to learn the trend feature of complicated nonstationary time series, a frequency-domain block is used to capture dynamic and complicated periodic patterns of time series data, and a fully connected neural network is used to combine the trend and periodic features for producing a final forecast. An adaptive frequency mechanism consisting of phase adaption, frequency adaption, and amplitude adaption is designed for mapping the frequency spectrum of the current sliding window to that of the forecasting interval. The multilayer neural networks conduct a transformation similar to the inverse discrete Fourier transform for generating a periodic feature forecast. Synthetic data and real-world data with different periodic characteristics are used to evaluate the effectiveness of the proposed model. The experimental results indicate that the ATFN has promising performance and strong adaptability for long-term time-series forecasting.

Fuzzy forecasting for long-term time series based on time-variant fuzzy information granules

As to the long-term time series forecasting, it is more challenging and practical to obtain the trend information and fluctuation range of sequence data than single-step prediction values. In this article, by means of fuzzy information granules (FIGs) and recurrent fuzzy neural networks, a novel long-term prediction model for time series is proposed. Based on a variable-length division method, generalized zonary time-variant fuzzy information granule (GZT-FIG) is constructed, which can express the variation trend, fluctuation range and dispersion degree of sequence data. Furthermore, in order to improve the anti-noise ability and memory ability, type-2 fuzzy sets and long short-term memory mechanism are introduced into the prediction scheme, based on which a self-evolving interval type-2 LSTM fuzzy neural network (eIT2FNN-LSTM) is provided. Compared with the existing works related to fuzzy-neural models, the involvement of long short-term memory mechanism effectively improves the memory ability to achieve the long-term prediction. In order to verify the validation and effectiveness of the proposed scheme, several groups of experiments, including synthetic sequences and real-life time series, are carried out. The experimental results reveal the better predictive performance and rich semantic information.

The practice of connectionist model for predicting forest fires in the Arctic zones of the Krasnoyarsk Territory

This paper presents data on the localization and number of forest fires that occur in the Arctic zones of the Krasnoyarsk Territory, as well as the possible causes of their occurrence. It is established that the main cause of forest and landscape fires are natural phenomena. A database of data normalized in a certain way about the factors shaping the occurrence of natural forest and landscape fires has been formed. The practice of connectionist algorithms, a forecasting model has been developed, and, based on data on forest and landscape fires in the Krasnoyarsk Territory; a model has been evaluated for using the model to predict fires in 2018. In order to compile a forecast of forest and landscape fires for 2019 in the Arctic zones of the Krasnoyarsk Territory using the connectionist algorithms, an optimal neuroarchitecture was chosen, which allows long-term time series forecasting.

Short and Long-Term Time Series Forecasting Stochastic Analysis for Slow Dynamic Processes

This paper intends to develop suitable methods to provide likely scenarios in order to support decision making for slow dynamic processes such as the underlying of agribusiness. A new method to analyze the short- and long-term time series forecast and to model the behavior of the underlying process using nonlinear artificial neural networks (ANN) is presented. The algorithm can effectively forecast the time-series data by stochastic analysis (Monte Carlo) of its future behavior using fractional Gaussian noise (fGn). The algorithm was used to forecast country risk time series for several countries, both for short term that is 30 days ahead and long term 350 days ahead scenarios.

Long-term forecasting of time series based on linear fuzzy information granules and fuzzy inference system

Long-term time series forecasting is a challenging problem both in theory and in practice. Although the idea of information granulation has been shown to be an essential concept and algorithmic pursuit in time series prediction, there is still an acute need for developing a sound conceptual framework for time series prediction so that information granulation can capture the essence of collections of data better, including average and trend information. In this paper, a novel type of fuzzy information granule involving a time-dependent (non-stationary) membership function is proposed to structure numerical time series into granular time series. We show that the underlying arithmetic along with the concept of distance for this type of information granules can be expressed in a simple way, which facilitates the ensuing processing of information granules. With this regard, distances between observation granules and antecedent granules presented in fuzzy rules can be easily determined. The design of long-term prediction method based on fuzzy inference system is then realized through interpolation completed with the aid of fuzzy rules. Experiments involving chaotic Mackey–Glass time series and real-world time series demonstrate that the proposed model produces better long-term forecasting than some existing numeric models such as Autoregressive (AR) models, nonlinear autoregressive (NAR) neural networks, Support Vector Regression (SVR) and fuzzy inference systems involving triangular and interval information granules.

Can we predict the unpredictable?

Time series forecasting is of fundamental importance for a variety of domains including the prediction of earthquakes, financial market prediction, and the prediction of epileptic seizures. We present an original approach that brings a novel perspective to the field of long-term time series forecasting. Nonlinear properties of a time series are evaluated and used for long-term predictions. We used financial time series, medical time series and climate time series to evaluate our method. The results we obtained show that the long-term prediction of complex nonlinear time series is no longer unrealistic. The new method has the ability to predict the long-term evolutionary trend of stock market time series, and it attained an accuracy level with 100% sensitivity and specificity for the prediction of epileptic seizures up to 17 minutes in advance based on data from 21 epileptic patients. Our new method also predicted the trend of increasing global temperature in the last 30 years with a high level of accuracy. Thus, our method for making long-term time series predictions is vastly superior to existing methods. We therefore believe that our proposed method has the potential to be applied to many other domains to generate accurate and useful long-term predictions.

Multiple-output modeling for multi-step-ahead time series forecasting

Accurate prediction of time series over long future horizons is the new frontier of forecasting. Conventional approaches to long-term time series forecasting rely either on iterated one-step-ahead predictors or direct predictors. In spite of their diversity, iterated and direct techniques for multi-step-ahead forecasting share a common feature, i.e. they model from data a multiple-input single-output mapping. In previous works, the authors presented an original multiple-output approach to multi-step-ahead prediction. The goal is to improve accuracy by preserving in the forecasted sequence the stochastic properties of the training series. This is not guaranteed for instance in direct approaches where predictions for different horizons are performed independently. This paper presents a review of single-output vs. multiple-output approaches for prediction and goes a step forward with respect to the previous authors contributions by (i) extending the multiple-output approach with a query-based criterion and (ii) presenting an assessment of single-output and multiple-output methods on the NN3 competition datasets. In particular, the experimental section shows that multiple-output approaches represent a competitive choice for tackling long-term forecasting tasks.