Proceeding Paper

Combining Forecasts of Time Series with Complex Seasonality

Using LSTM-Based Meta-Learning †
Grzegorz Dudek

Electrical Engineering Faculty, Czestochowa University of Technology, Al. AK 17, 42-200 Cz˛estochowa, Poland;
[email protected]
† Presented at the 9th International Conference on Time Series and Forecasting, Gran Canaria, Spain,
12–14 July 2023.

Abstract: In this paper, we propose a method for combining forecasts generated by different models
based on long short-term memory (LSTM) ensemble learning. While typical approaches for combining
forecasts involve simple averaging or linear combinations of individual forecasts, machine learning
techniques enable more sophisticated methods of combining forecasts through meta-learning, leading
to improved forecasting accuracy. LSTM’s recurrent architecture and internal states offer enhanced
possibilities for combining forecasts by incorporating additional information from the recent past. We
define various meta-learning variants for seasonal time series and evaluate the LSTM meta-learner
on multiple forecasting problems, demonstrating its superior performance compared to simple
averaging and linear regression.

Keywords: ensemble forecasting; LSTM; machine learning; multiple seasonal patterns; short-term
load forecasting

1. Introduction
Real-world time series can exhibit various complex properties such as time-varying
trends, multiple seasonal patterns, random fluctuations, and structural breaks. Given this
complexity, it can be challenging to identify a single best model to accurately approximate
the underlying data-generating process [1]. To address this issue, a common approach is to
Citation: Dudek, G. Combining
combine multiple forecasting models to capture the multiple drivers of the data-generating
Forecasts of Time Series with
process and mitigate uncertainties regarding model form and parameter specification [2].
Complex Seasonality Using
This approach, known as ensemble forecasting or combining forecasts, has been shown to
LSTM-Based Meta-Learning. Eng.
be effective in improving the accuracy and reliability of time series forecasts. By combining
Proc. 2023, 39, 53. https://doi.org/
10.3390/engproc2023039053
forecasts, the aim is to take advantage of the strengths of multiple models and reduce the
impact of their individual weaknesses.
Academic Editors: Ignacio Rojas, There are several potential explanations for the strong performance of forecast combi-
Hector Pomares, Luis Javier Herrera, nations. Firstly, by combining forecasts, the resulting ensemble can capture a broader range
Fernando Rojas and Olga Valenzuela
of information and better handle the forecasting problem complexity. It can leverage the
Published: 5 July 2023 strengths of individual models, as each model may capture different aspects of the under-
lying data-generating process. Therefore, the resulting ensemble can incorporate partial
and incompletely overlapping information, leading to improved accuracy and robustness.
Secondly, in the presence of structural breaks and other instabilities, combining forecasts
Copyright: © 2023 by the author. from models with different degrees of misspecification and adaptability can mitigate the
Licensee MDPI, Basel, Switzerland. problem. This is because individual models may perform well under certain conditions but
This article is an open access article
poorly under others, and by combining them, the ensemble can better handle a range of
distributed under the terms and
potential scenarios [3]. Finally, forecast combinations can improve stability compared to
conditions of the Creative Commons
using a single model, as the ensemble is less sensitive to the idiosyncrasies of individual
Attribution (CC BY) license (https://
models. This means that the resulting forecasts are less likely to be influenced by outliers
creativecommons.org/licenses/by/
or errors in individual models, leading to more reliable predictions.
4.0/).

Eng. Proc. 2023, 39, 53. https://doi.org/10.3390/engproc2023039053 https://www.mdpi.com/journal/engproc

Eng. Proc. 2023, 39, 53 2 of 10

In a classical way, by combining the predictions from multiple models, the resulting
ensemble prediction can be thought of as an average of the individual predictions. The vari-
ance of the average of multiple independent random variables is typically lower than the
variance of a single random variable, assuming that the individual predictions are diverse.
Therefore, a key issue in ensemble learning is ensuring diversity among the individual
models being combined. If the models are too similar, the ensemble may not be able to
capture the full range of possible outcomes and may not improve predictive performance.
In this work, we ensure high diversity among models by using non-interfering models
with different operating principles and architectures, including statistical, machine learning
(ML), and hybrid models (see Section 3.2).
A simple arithmetic average of forecasts based on equal weights is a popular and
surprisingly robust combination rule, outperforming more complicated weighting schemes
in many cases [4,5]. Other strategies, such as using the median, mode, trimmed means,
and winsorized means, are also applied [6]. To differentiate weights assigned to individual
models, linear regression can be used, where the vector of past observations is the response
variable and the matrix of past individual forecasts is the predictor variable. Combination
weights can be estimated using ordinary least squares. The weights can reflect individual
models’ performance on historical data [7]. Time-varying weights can be used to improve
forecasting ability in the presence of instabilities, and principal components regression can
be used as a solution for multicollinearity [8]. Weights can also be derived from information
criteria such as AIC [9].
Linear combination approaches assume a linear dependence between constituent
forecasts and the variable of interest, and may not result in the best forecast, especially if
the individual forecasts come from nonlinear models or if the true relationship between
base forecasts and the target has a nonlinear form [10]. In contrast, ML models can combine
the base forecasts nonlinearly using a stacking procedure.
Stacking is an ensemble ML algorithm that learns how to best combine predictions
from multiple models, using the concept of meta-learning to boost forecasting accuracy
beyond that achieved by the individual models. Neural networks (NNs) are often used
in stacking to estimate the nonlinear mapping between the target value and its forecasts
produced by multiple models [11]. The power of ensemble learning for forecasting was
demonstrated in [12], where several meta-learning approaches were evaluated on a large
and diverse set of time series data. Ensemble methods were found to provide a benefit
in overall forecasting accuracy, with simple ensemble methods leading to good results on
average. However, there was no single meta-learning method that was suitable for all
time series.
The main contributions of this study can be summarized in the following three aspects:
1. A meta-learning approach based on LSTM is proposed for combining forecasts. This
approach incorporates past information accumulated in the internal states, improving
accuracy, especially in cases where there is a temporal relationship between base
forecasts for successive time points.
2. Various meta-learning variants for time series with multiple seasonal patterns are
proposed, such as the use of the full training set, including base forecasts for successive
time points, and the use of selected training points that reflect the seasonal structure
of the data.
3. Extensive experiments are conducted on 35 time series with triple seasonality us-
ing 16 base models to validate the efficacy of the proposed approach. The exper-
imental results demonstrate the high performance of the LSTM meta-learner and
its potential to combine forecasts more accurately than simple averaging and linear
regression methods.
The remainder of this work is structured as follows. Section 2 presents the pro-
posed LSTM meta-model and introduces both the global and local meta-learning variants.
Section 3 provides application examples for time series with complex seasonality and
Eng. Proc. 2023, 39, 53 3 of 10

discusses the results obtained from the conducted experiments. Finally, in Section 4, we
conclude our work by summarizing the key findings and contributions.

2. LSTM for Combining Forecasts

The problem of forecast combinations refers to the task of finding regression function f
that aggregates the forecasts for time t produced by n forecasting models. The function can
use all the available information up to time t − h, where h is a forecast horizon, but in this
study, we limit this information to the base forecasts expressed by vector ŷt = [ŷ1,t , ..., ŷn,t ].
The combined forecast is ỹt = f (ŷt ; θt ), where θt is a vector of meta-model parameters.
The model learns using training set Φ = {ŷτ , yτ }τ ∈Ξ , where yτ is a target value and
Ξ is a set of selected time indexes from interval T = 1, ..., t − h (selection of this set is
considered in Section 2.2).
The class of regression functions f encompasses both linear and nonlinear mappings,
as well as series-specific and cross-learning mappings. In the latter approach, the parame-
ters of the function are selected through a learning process over multiple time series, which
enhances the generalization capability of the model. Furthermore, the parameters can either
be static or time-varying throughout the forecasting horizon. To maximize the performance
of the ensemble, we adopt an approach where we learn the meta-model parameters for each
forecasting task individually, using a specific training set for each task (see Section 2.2).

2.1. LSTM Model

LSTM is a modern recurrent NN that incorporates gating mechanisms [13]. This NN
architecture was specifically designed to handle sequential data and is capable of learning
short and long-term relationships in time series [14]. LSTM is composed of recurrent cells
that can maintain their internal states over time, i.e., cell state c and hidden state h. These
cells are regulated by nonlinear gating mechanisms that control the flow of information
within the cell, allowing it to adapt to the dynamics of the current process.
In our implementation, the LSTM network consists of two layers: the LSTM layer and
the linear layer, see Figure 1. The LSTM layer is responsible for approximating temporal
nonlinear dependencies in sequential data and generating state vectors. On the other hand,
the linear layer converts hidden state vector h into the output value. The aggregation
function implemented in the LSTM network can be written as:

f (ŷt ) = vT ht (ŷt ) + v0 (1)

ht (ŷt ) = LSTM(ŷt , ct−1 , ht−1 ; w) ∈ Rm (2)

where w and v are the weights of the LSTM and linear layers, respectively.

Linear layer

Figure 1. LSTM model.

Eng. Proc. 2023, 39, 53 4 of 10

The number of nodes in each gate, m, is the most critical hyperparameter. It determines
the amount of information stored in the states. For more intricate temporal relationships, a
higher number of nodes is necessary.
In contrast to non-recurrent ML models such as feed-forward NNs, tree-based models,
and support vector regression, to calculate output ỹt , LSTM uses not only the information
included in the base forecasts for time t, ŷt , but also in the base forecasts for previous time
steps, t − 1, t − 2, . . .. This is achieved through states ct−1 and ht−1 , which accumulate
information from the past steps.

2.2. Meta-Learning Variants

The forecasting models generate forecasts for the successive time points T = 1, . . . , t − h.
To obtain an ensemble forecast for time t, we can train the meta-model using all available
data from the historical period, i.e., Ξ = T, which is referred to as the global approach.
Using this method, the model can utilize all available information to generate a forecast for
the current time point.
In local learning, we restrict the training sequence to the last k points, i.e.,
Ξ = t − h − k, . . . , t − h, allowing the LSTM to model the relationship for the query pattern
ŷt based on the most recent sequence of length k. We refer to this approach as v1.
When ensembling seasonal time series, training the LSTM model on points from the
same phase of the cycle as the forecasted point can improve forecast accuracy. In this
approach, the training set consists of points Ξ = {t − ks1 , t − (k − 1)s1 , . . . , t − s1 }, where
s1 denotes the period of the seasonal cycle and k is a predefined size of the training set. It is
worth noting that this training set retains the time structure of the data, but simplifies it
by only including points that are in the same phase of the seasonal cycle as the forecasted
point. We refer to this approach as v2.
In the case of double seasonality with periods s1 and s2 (assuming that s2 is a multiple
of s1 ), we can create the training set by selecting points from the same phase of both
seasonal patterns as the forecasted point. Specifically, the training set is composed of points
Ξ = {t − ks2 , t − (k − 1)s2 , . . . , t − s2 }. We refer to this approach as v3. Figure 2 visualizes
the training target points for each variant of LSTM learning.

s1
s2

Real v1 v2 v3 Forecasted point

Figure 2. Selection of training points for LSTM.

Note that approaches v2 and v3 remove the training points that are not in the same
phase as the forecasted point. This simplifies the relationship between the new training
points and the forecasted point, making it easier to model. However, this simplification
comes at the cost of potentially losing some of the information related to the seasonal
patterns that occur outside of the selected phase. Therefore, it is important to carefully
consider which approach to use depending on the specific characteristics of the data.

3. Experimental Study
We evaluate the performance of our proposed approach, combining forecasts gen-
erated by 16 forecasting models described in Section 3.2. The forecasting problem is
short-term load forecasting for 35 European countries.
Eng. Proc. 2023, 39, 53 5 of 10

3.1. Data, Forecasting Problem and Research Design

We use the real-world data collected from the ENTSO-E repository (www.entsoe.eu/
data/power-stats accessed on 6 April 2016). The dataset includes hourly electricity loads
spanning from 2006 to 2018, representing 35 European countries. It offers a diverse set of
time series, each exhibiting unique properties such as distinct levels and trends, variance
stability over time, intensity and regularity of seasonal fluctuations spanning different
periods (annual, weekly, and daily), and varying degrees of random fluctuations.
The forecasting models were optimized using data from 2006 to 2017 and applied to
generate hourly forecasts for the year 2018, day by day. To evaluate the performance of the
combining model, 100 hours for each country were chosen from the second half of 2018
(evenly spaced across the period) and the forecasts for each of these hours were combined
using LSTM. The LSTM model was trained separately for each selected hour, with preceding
data spanning from 1 January 2018 up to the hour preceding the forecasted hour (h = 1)
used for optimization and training across three variants (v1, v2, and v3). This resulted in a
total of 10,500 training sessions (35 · 100 · 3). In variant v2, we assumed daily seasonality
period s1 = 24 h, while in variant v3 we assumed weekly period s2 = 7 · 24 = 168 h.
This study utilized Matlab implementation of the LSTM model. Some LSTM hyperpa-
rameters were set to default values, while others were determined through experimentation.
The latter include the number of nodes m = 128, and the number of epochs—200.
As performance metrics, the following measures were used: MAPE—mean absolute
percentage error, MdAPE—median of absolute percentage error, MSE—mean square error,
MPE—mean absolute percentage error, and StdPE—standard deviation of percentage error.

3.2. Forecasting Models

As the base forecasting models, we use a set of statistical models and classical ML
models, as well as recurrent, deep, and hybrid NN architectures from [15]:
• ARIMA—auto-regressive integrated moving average model,
• ETS—exponential smoothing model,
• Prophet—modular additive regression model with nonlinear trend and seasonal
components,
• N-WE—Nadaraya–Watson estimator,
• GRNN—general regression NN,
• MLP—perceptron with a single hidden layer and sigmoid nonlinearities,
• SVM—linear epsilon insensitive support vector machine (e-SVM),
• LSTM—long short-term memory,
• ANFIS—adaptive neuro-fuzzy inference system,
• MTGNN—graph NN for multivariate time series forecasting,
• DeepAR—autoregressive recurrent NN model for probabilistic forecasting,
• WaveNet—autoregressive deep NN model combining causal filters with dilated
convolutions,
• N-BEATS—deep NN with hierarchical doubly residual topology,
• LGBM—Light Gradient-Boosting Machine,
• XGB—eXtreme Gradient-Boosting algorithm,
• cES-adRNN—contextually enhanced hybrid and hierarchical model combining ETS
and dilated RNN with attention mechanism.

3.3. Results and Discussion

Table 1 shows the forecasting quality metrics for the base forecasting models. Note the
significant difference in results between the various models, with MAPE ranging from 1.70
for cES-adRNN to 3.83 for Prophet. The overall mean MAPE across all models was 2.53.
Table 2 shows forecasting quality metrics for different ensemble approaches. Mean
and Median are just the mean and median of 16 forecasts produced by the base models.
LinReg is a linear combination of these forecasts with weights estimated on the training
samples Ξ = T. As can be seen from Table 2, the most accurate approach is variant v1
Eng. Proc. 2023, 39, 53 6 of 10

of LSTM for k = 168. This variant, which involves meta-learning on the full sequence
restricted to the last 168 points, provided the most accurate results as measured by MAPE,
MdAPE, and MSE errors. Note the significant difference in errors between this variant and
the second most accurate ensembling method, LinReg, which achieved about 5% in MAPE
and 35% in MSE.

Table 1. Forecasting quality metrics for the base models.

MAPE MdAPE MSE MPE StdPE

ARIMA 2.86 1.82 777,012 0.0556 4.60
ETS 2.83 1.79 710,773 0.1639 4.64
Prophet 3.83 2.53 1,641,288 −0.5195 6.24
N-WE 2.12 1.34 357,253 0.0048 3.47
GRNN 2.10 1.36 372,446 0.0098 3.42
MLP 2.55 1.66 488,826 0.2390 3.93
SVM 2.16 1.33 356,393 0.0293 3.55
LSTM 2.37 1.54 477,008 0.0385 3.68
ANFIS 3.08 1.65 801,710 −0.0575 5.59
MTGNN 2.54 1.71 434,405 0.0952 3.87
DeepAR 2.93 2.00 891,663 −0.3321 4.62
WaveNet 2.47 1.69 523,273 −0.8804 3.77
N-BEATS 2.14 1.34 430,732 −0.0060 3.57
LGBM 2.43 1.70 409,062 0.0528 3.55
XGB 2.32 1.61 376,376 0.0529 3.37
cES-adRNN 1.70 1.10 224,265 −0.1860 2.57

Note that using the simplest method of combining forecasts, Mean or Median, resulted
in significantly larger errors compared to LSTM v1. Unfortunately, variants v2 and v3,
which excluded seasonality from the training sequence, were found to be inaccurate and
did not perform well. This suggests that excluding seasonality from the training sequence
could lead to the loss of important information related to the seasonal patterns in the data,
resulting in deteriorated forecasting performance.
Figure 3 displays the MAPE boxplots for LSTM in three variants with varying lengths
of the training sequence k. Additionally, the boxplots for the baseline methods, namely
Mean, Median, and LinReg, are shown for comparison. As shown in the figure, LSTM in
variants v2 and v3 are highly sensitive to the length of the training sequence. It achieved
the lowest errors when trained on all available data points. Extending the training sequence
may potentially further reduce errors. In contrast, for LSTM v1, the training sequences of
length 168 hours (one week) provided the lowest errors.
MPE in Table 2 provides information about the forecast bias, which is the lowest
for LinReg, but LSTM v1, with MPE = 0.0247, is in second place. It is worth noting
that Mean and Median produce more biased forecasts. The lowest value of StdPE for
LSTM v1 indicates the least dispersed predictions compared to other approaches for
combining forecasts.

Table 2. Forecasting quality metrics for different ensemble approaches (best results in bold).

Variant MAPE MdAPE MSE MPE StdPE

Mean - 1.91 1.23 316,943 −0.0775 3.11
Median - 1.82 1.13 287,284 −0.0682 3.05
LinReg - 1.63 1.11 213,428 0.0131 2.38
LSTM v1, k = 168 1.55 1.09 139,667 0.0247 2.26
LSTM v2, global 1.95 1.34 270,266 −0.1046 2.89
LSTM v3, global 2.97 1.84 726,108 −0.3628 4.84
Eng. Proc. 2023, 39, 53 7 of 10

Figure 3. MAPE boxplots for the various ensemble variants.

Figure 4 depicts examples of forecasts for selected countries and test points. It is worth
noting that LSTM v1 was able to achieve forecasts close to the target values, which were
outside the interval of the base models’ forecasts (let us denote this interval for the i-th
test point by Zi ) and despite the fact that no base model even came close to these targets
(see test point no. 94 for FR and 99 for GB in Figure 4). One possible explanation for this
ability of LSTM is the incorporation of additional information from the immediate past
through internal states c and h (see (2)). LinReg, having no internal states, cannot use such
information. Mean and Median approaches cannot even go beyond the interval Zi .
To test the ability of LSTM v1 and LinReg to produce forecasts outside the interval Zi ,
we counted the number of such cases out of the 3500 forecasts produced by each model.
The results are shown in column N1 of Table 3. Column N2 counts how many of these N1
cases concern the situation where the target value also lay outside the Z-interval, on the
same side as the meta-model forecast. Column N3 counts the number of cases out of N1 for
which the meta-model produces more accurate predictions than the Median approach. It is
evident from Table 3 that LSTM generates many more forecasts outside of Zi than LinReg.
This may indicate better extrapolation properties of LSTM, but on the other hand, it may
also suggest an increased susceptibility to overfitting.

Table 3. Extrapolation properties of LSTM v1 and LinReg.

N1 N2 N3
LinReg 48 13 27
LSTM v1 447 192 244

In summary, our research findings suggest that LSTM, as a meta-learner, exhibits

sensitivity to the length of the training sequence and achieves optimal performance when
trained in global mode. However, it is important to note that the overall performance
also depends on the accuracy and correlation of the base forecasts. In this study, we did
not delve into the analysis of interdependence between the base forecasts or select the
optimal set of base models. These aspects present opportunities for further optimization
and improvement in future research.
LSTM poses greater challenges compared to classical ML methods such as MLP or
random forests. It involves a larger number of hyperparameters and parameters that need
to be tuned, making the optimization and training process more complex. Additionally,
LSTM typically requires a larger amount of data to achieve optimal performance due to
its ability to capture intricate temporal dependencies. In local versions of training, where
shorter training sequences are used, accurate predictions with LSTM can be challenging
to obtain. This highlights the importance of having sufficient training data to effectively
capture the underlying patterns and dynamics of the sequential data.
Eng. Proc. 2023, 39, 53 8 of 10

104 FR Base forecast LSTM, v1

104 ES
Mean LSTM, v2
8.5 Median LSTM, v3 3.6
LinReg Real

8 3.4

3.2
7.5

3
7

y
2.8
6.5
2.6
6
2.4

5.5 2.2

5 2
80 85 90 95 100 20 25 30 35 40
Test point Test point

104 GB 104 DE
6 8.5

8
5.5
7.5
5
7

4.5 6.5
y

y
4 6

5.5
3.5
5
3
4.5

2.5 4
80 85 90 95 100 80 85 90 95 100
Test point Test point

Figure 4. Base and ensemble forecasts.

4. Conclusions
This study proposes a meta-learning approach for combining forecasts based on LSTM,
which has the potential to improve accuracy, particularly in cases where there is a temporal
relationship between base forecasts. The study also proposes different variants of the
approach for time series with multiple seasonal patterns.
The experimental results clearly demonstrate that the LSTM meta-learner outperforms
simple averaging, median, and linear regression methods in terms of forecasting accuracy.
In addition, LSTM has distinct advantages over non-recurrent ML models as it is capable of
leveraging its internal states to model dependencies between base forecasts for consecutive
time points and capture patterns in the sequential data.
Further studies could compare LSTM with other meta-learning approaches, such
as feed-forward and randomized NNs, random forests, and boosted trees to determine
which approach is best suited for a given forecasting problem. Moreover, selecting a
pool of base models and controlling their diversity is an interesting topic that requires
further investigation.

Funding: This researcher was supported by grant 020/RID/2018/19 “Regional Initiative of Excel-
lence” from the Polish Minister of Science and Higher Education, 2019–23.
Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.
Data Availability Statement: We use real-world data collected from www.entsoe.eu (accessed on 6
April 2016).
Acknowledgments: The author thanks Slawek Smyl and Paweł Pełka for providing forecasts from
the base models.
Conflicts of Interest: The author declares no conflict of interest.
Eng. Proc. 2023, 39, 53 9 of 10

Abbreviations
The following abbreviations are used in this manuscript:
ANFIS Adaptive Neuro-Fuzzy Inference System
ARIMA Auto-Regressive Integrated Moving Average
cES-adRNN contextually enhanced hybrid and hierarchical model combining ETS and dilated
RNN with attention mechanism
DE Germany
DeepAR Auto-Regressive Deep recurrent NN model for probabilistic forecasting
ES Spain
ETS Exponential Smoothing
FR France
GB Great Britain
GRNN General Regression Neural Network
LinReg Linear Regression
LGBM Light Gradient-Boosting Machine
LSTM Long Short-Term Memory Neural Network
MAPE Mean Absolute Percentage Error
MdAPE Median of Absolute Percentage Error
ML Machine Learning
MLP Multilayer Perceptron
MPE Mean Percentage Error
MSE Mean Square Error
MTGNN Graph Neural Network for Multivariate Time series forecasting
N-BEATS deep NN with hierarchical doubly residual topology
N-WE Nadaraya—Watson Estimator
NN Neural Network
PE Percentage Error
PL Poland
RNN Recurrent Neural Network
StdPE Standard Deviation of Percentage Error
SVM Support Vector Machine
STLF Short-Term Load Forecasting
WaveNet Auto-Regressive deep NN model combining causal filters with dilated convolutions
XGB eXtreme Gradient Boosting

References
1. Clements, M.; Hendry, D. Forecasting Economic Time Series; Cambridge University Press: Cambridge, UK, 1998.
2. Wang, X.; Hyndman, R.; Li, F.; Kang, Y. Forecast combinations: An over 50-year review. Int. J. Forecast. 2022, in press. [CrossRef]
3. Rossi, B. Forecasting in the presence of instabilities: How we know whether models predict well and how to improve them.
J. Econ. Lit. 2021, 59, 1135–1190. [CrossRef]
4. Blanc, S.; Setzer, T. When to choose the simple average in forecast combination. J. Bus. Res. 2016, 69, 3951–3962. [CrossRef]
5. Genre, V.; Kenny, G.; Meyler, A.; Timmermann, A. Combining expert forecasts: Can anything beat the simple average? Int. J.
Forecast. 2013, 29, 108–121. [CrossRef]
6. Jose, V.; Winkler, R. Simple robust averages of forecasts: Some empirical results. Int. J. Forecast. 2008, 24, 163–169. [CrossRef]
7. Pawlikowski, M.; Chorowska, A. Weighted ensemble of statistical models. Int. J. Forecast. 2020, 36, 93–97. [CrossRef]
8. Poncela, P.; Rodriguez, J.; Sanchez-Mangas, R.; Senra, E. Forecast combination through dimension reduction techniques. Int. J.
Forecast. 2011, 27, 224–237. [CrossRef]
9. Kolassa, S. Combining exponential smoothing forecasts using Akaike weights. Int. J. Forecast. 2011, 27, 238–251. [CrossRef]
10. Babikir, A.; Mwambi, H. Evaluating the combined forecasts of the dynamic factor model and the artificial neural network model
using linear and nonlinear combining methods. Empir. Econ. 2016, 51, 1541–1556. [CrossRef]
11. Zhao, S.; Feng, Y. For2For: Learning to forecast from forecasts. arXiv 2020, arXiv:2001.04601.
12. Gastinger, J.; Nicolas, S.; Stepić, D.; Schmidt, M.; Schülke, A. A study on ensemble learning for time series forecasting and the
need for meta-learning. In Proceedings of the 2021 International Joint Conference on Neural Networks (IJCNN), Shenzhen, China,
18–22 July 2021; pp. 1–8.
13. Hochreiter S.; Schmidhuber J. Long short-term memory. Neural Comput. 1997, 9, 1735–1780. [CrossRef] [PubMed]
Eng. Proc. 2023, 39, 53 10 of 10

14. Hewamalage H.; Bergmeir C.; Bandara K. Recurrent neural networks for time series forecasting: Current status and future
directions. Int. J. Forecast. 2021, 37, 388–427. [CrossRef]
15. Smyl, S.; Dudek, G.; Pełka, P. Contextually enhanced ES-dRNN with dynamic attention for short-term load forecasting. arXiv
2022, arXiv:2212.09030.

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual
author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to
people or property resulting from any ideas, methods, instructions or products referred to in the content.