|
on Forecasting |
| By: | McCloskey, PJ; Malheiros Remor, Rodrigo |
| Abstract: | Forecasting GDP is crucial for economic planning and policymaking. This study compares the performance of three widely-used econometric models—ARIMA, VAR, and Linear Regression—using GDP data from the UAE. Employing a rolling forecast approach, we analyze the models’ accuracy over different time horizons. Results indicate ARIMA’s robust long-term forecasting capability, LR models perform better with short-term predictions, particularly when exogenous variable forecasts are accurate. These insights provide a valuable foundation for selecting forecasting models in the UAE’s evolving economy, suggesting ARIMA’s suitability for long-term outlooks and LR for short-term, scenario-based forecasts. |
| Keywords: | GDP forecasting, ARIMA, VAR, Linear Regression, UAE economy |
| JEL: | O1 O10 O4 O40 |
| Date: | 2024–09–10 |
| URL: | https://d.repec.org/n?u=RePEc:pra:mprapa:122860 |
| By: | Xiaoqian Wang; Rob J Hyndman; Shanika Wickramasuriya |
| Abstract: | Forecast reconciliation ensures forecasts of time series in a hierarchy adhere to aggregation constraints, enabling aligned decision making. While forecast reconciliation can enhance overall accuracy in hierarchical or grouped structures, the most substantial improvements occur  in series with initially poor-performing base forecasts. Nevertheless, certain series may experience deteriorations in reconciled forecasts. In practical settings, series in a structure often exhibit poor base forecasts due to model  misspecification or low forecastability. To prevent their negative impact, we propose two categories of forecast reconciliation methods that incorporate time series selection based on  out-of-sample and in-sample information, respectively. These methods keep “poor†base forecasts unused in forming reconciled forecasts, while adjusting weights allocated to the remaining series accordingly when generating bottom-level reconciled forecasts. Additionally, our methods ameliorate disparities stemming from varied estimates of the base forecast error covariance matrix, alleviating challenges associated with estimator selection. Empirical evaluations through two simulation studies and applications using Australian labour force and domestic tourism data demonstrate improved forecast accuracy, particularly evident in higher aggregation levels, longer forecast horizons, and cases involving model misspecification. |
| Keywords: | Forecasting; Hierarchical Time Series; Grouped Time Series; Linear Forecast Reconciliation; Integer Programming |
| Date: | 2024 |
| URL: | https://d.repec.org/n?u=RePEc:msh:ebswps:2024-5 |
| By: | Xiaoqian Wang; Rob J Hyndman |
| Abstract: | We consider the problem of constructing distribution-free prediction intervals for multi-step​ time series forecasting, with a focus on the temporal dependencies inherent in multi-step forecast​ errors. We establish that the optimal h-step-ahead forecast errors exhibit serial correlation​ up to lag (h-1) under a general non-stationary autoregressive data generating process. To​ leverage these properties, we propose the Autocorrelated Multi-step Conformal Prediction(AcMCP) method, which effectively incorporates autocorrelations in multi-step forecast errors, resulting in more statistically efficient prediction intervals. This method ensures theoretical​ long-run coverage guarantees for multi-step prediction intervals, though we note that increasedf​ orecasting horizons may exacerbate deviations from the target coverage, particularly in the​ context of limited sample sizes. Additionally, we extend several easy-to-implement conformal prediction methods, originally designed for single-step forecasting, to accommodate multi-step scenarios. Through empirical evaluations, including simulations and applications to data, we demonstrate that AcMCP achieves coverage that closely aligns with the target within local windows, while providing adaptive prediction intervals that effectively respond to varying conditions. |
| Keywords: | Conformal Prediction, Coverage Guarantee, Distribution-Free Inference, Exchangeability, Weighted Quantile Estimate |
| Date: | 2024 |
| URL: | https://d.repec.org/n?u=RePEc:msh:ebswps:2024-20 |
| By: | Paul Ghelasi; Florian Ziel |
| Abstract: | Power prices can be forecasted using data-driven models or fundamental models. Data-driven models learn from historical patterns, while fundamental models simulate electricity markets. Traditionally, fundamental models have been too computationally demanding to allow for intrinsic parameter estimation or frequent updates, which are essential for short-term forecasting. In this paper, we propose a novel data-driven fundamental model that combines the strengths of both approaches. We estimate the parameters of a fully fundamental merit order model using historical data, similar to how data-driven models work. This removes the need for fixed technical parameters or expert assumptions, allowing most parameters to be calibrated directly to observations. The model is efficient enough for quick parameter estimation and forecast generation. We apply it to forecast German day-ahead electricity prices and demonstrate that it outperforms both classical fundamental and purely data-driven models. The hybrid model effectively captures price volatility and sequential price clusters, which are becoming increasingly important with the expansion of renewable energy sources. It also provides valuable insights, such as fuel switches, marginal power plant contributions, estimated parameters, dispatched plants, and power generation. |
| Date: | 2025–01 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2501.02963 |
| By: | Yangzhuoran Fin Yang; Rob J Hyndman; George Athanasopoulos; Anastasios Panagiotelis |
| Abstract: | We propose a novel forecast linear augmented projection (FLAP) method that can reduce the forecasterror variance of any multivariate forecast. The method first constructs new component series whichare linear combinations of the original series. Forecasts are then generated for both the original andcomponent series. Finally, the full vector of forecasts is projected onto a linear subspace where theconstraints implied by the combination weights hold. We show that the projection using the originalforecast error covariance matrix will result in improved forecasts. Notably, the new forecast errorvariance of each series is non-increasing with the number of components, and mild conditions are established for which it is strictly decreasing. It is also shown that the proposed method achieves maximum forecast error variance reduction among linear projection methods. We demonstrateour proposed method with an estimated covariance matrix using simulations and two empirical applications based on Australian tourism and FRED-MD data. In all cases,  forecasts are improved. Notably, using FLAP with Principal Component Analysis (PCA) to construct the new series leads tosubstantial forecast error variance reduction. |
| Keywords: | Forecasting; Hierarchical time series; Grouped time series; Linear forecast reconciliation; Integer programming |
| Date: | 2024 |
| URL: | https://d.repec.org/n?u=RePEc:msh:ebswps:2024-13 |
| By: | Meg Tulloch |
| Abstract: | The primary objective of this dissertation is to improve the accuracy of the National Institute of Economic and Social Research (NIESR) forecasting model for predicting UK wage growth, as featured in their monthly Wage Tracker. This study investigates the application of several time series methodologies, inclusive of Auto-Regressive Integrated Moving Average (ARIMA), Vector Auto-Regression (VAR), Vector Error Correction Model (VECM), and Dynamic Factor Model (DFM), to enhance forecast precision and reduce uncertainty. By evaluating these models, the research aims to provide a more reliable framework for forecasting wage growth and better inform economic analysis and policy decisions. |
| Keywords: | Wages, Forecasting, Nowcasting |
| URL: | https://d.repec.org/n?u=RePEc:nsr:niesrd:565 |
| By: | Nuwani K Palihawadana; Rob J Hyndman; Xiaoqian Wang |
| Abstract: | Forecasting often involves high-dimensional predictors which have nonlinear relationships with theoutcome of interest. Nonparametric additive index models can capture these relationships, while addressing the curse of dimensionality. This paper introduces a new algorithm, Sparse Multiple Index(SMI) Modelling, tailored for estimating high-dimensional nonparametric/semi-parametric additive index models, while limiting the number of parameters to estimate, by optimising predictor selectionand predictor grouping. The SMI Modelling algorithm uses an iterative approach based on mixed integer programming to solve an ℓ0-regularised nonlinear least squares optimisation problem withlinear constraints. We demonstrate the performance of the proposed algorithm through a simulationstudy, along with two empirical applications to forecast heat-related daily mortality and daily solarintensity. |
| Keywords: | Additive Index Models; Variable Selection; Dimension Reduction; Predictor Grouping; Mixed Integer Programming |
| Date: | 2024 |
| URL: | https://d.repec.org/n?u=RePEc:msh:ebswps:2024-16 |
| By: | Chang, Jinyuan; Fang, Qin; Qiao, Xinghao; Yao, Qiwei |
| Abstract: | We propose a two-step procedure to model and predict high-dimensional functional time series, where the number of function-valued time series p is large in relation to the length of time series n. Our first step performs an eigenanalysis of a positive definite matrix, which leads to a one-to-one linear transformation for the original high-dimensional functional time series, and the transformed curve series can be segmented into several groups such that any two subseries from any two different groups are uncorrelated both contemporaneously and serially. Consequently in our second step those groups are handled separately without the information loss on the overall linear dynamic structure. The second step is devoted to establishing a finite-dimensional dynamical structure for all the transformed functional time series within each group. Furthermore the finite-dimensional structure is represented by that of a vector time series. Modeling and forecasting for the original high-dimensional functional time series are realized via those for the vector time series in all the groups. We investigate the theoretical properties of our proposed methods, and illustrate the finite-sample performance through both extensive simulation and two real datasets. Supplementary materials for this article are available online, including a standardized description of the materials available for reproducing the work. |
| Keywords: | dimension reduction; Eigenanalysis; functional thresholding; Hilbert–Schmidt norm; permutation; segmentation transformation |
| JEL: | C1 |
| Date: | 2024–11–26 |
| URL: | https://d.repec.org/n?u=RePEc:ehl:lserod:125599 |
| By: | Guhan Sivakumar |
| Abstract: | This paper explores the application of Hidden Markov Models (HMM) and Long Short-Term Memory (LSTM) neural networks for economic forecasting, focusing on predicting CPI inflation rates. The study explores a new approach that integrates HMM-derived hidden states and means as additional features for LSTM modeling, aiming to enhance the interpretability and predictive performance of the models. The research begins with data collection and preprocessing, followed by the implementation of the HMM to identify hidden states representing distinct economic conditions. Subsequently, LSTM models are trained using the original and augmented data sets, allowing for comparative analysis and evaluation. The results demonstrate that incorporating HMM-derived data improves the predictive accuracy of LSTM models, particularly in capturing complex temporal patterns and mitigating the impact of volatile economic conditions. Additionally, the paper discusses the implementation of Integrated Gradients for model interpretability and provides insights into the economic dynamics reflected in the forecasting outcomes. |
| Date: | 2025–01 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2501.02002 |
| By: | Jairo Flores (Banco Central de Reserva del Perú); Bruno Gonzaga (Banco Central de Reserva del Perú); Walter Ruelas-Huanca (Banco Central de Reserva del Perú); Juan Tang (Banco Central de Reserva del Perú) |
| Abstract: | This paper explores the application of machine learning (ML) techniques to nowcast the monthly year-over-year growth rate of both total and non-primary GDP in Peru. Using a comprehensive dataset that includes over 170 domestic and international predictors, we assess the predictive performance of 12 ML models, including Lasso, Ridge, Elastic Net, Support Vector Regression, Random Forest, XGBoost, and Neural Networks. The study compares these ML approaches against the traditional Dynamic Factor Model (DFM), which serves as the benchmark for nowcasting in economic research. We treat specific configurations, such as the feature matrix rotations and the dimensionality reduction technique, as hyperparameters that are optimized iteratively by the Tree-Structured Parzen Estimator. Our results show that ML models outperformed DFM in nowcasting total GDP, and that they achieve similar performance to this benchmark in nowcasting non-primary GDP. Furthermore, the bottom-up approach appears to be the most effective practice for nowcasting economic activity, as aggregating sectoral predictions improves the precision of ML methods. The findings indicate that ML models offer a viable and competitive alternative to traditional nowcasting methods. |
| Keywords: | GDP, Machine Learning, nowcasting |
| JEL: | C14 C32 E32 E52 |
| Date: | 2024–12 |
| URL: | https://d.repec.org/n?u=RePEc:rbp:wpaper:2024-019 |
| By: | Katsafados, Apostolos G.; Leledakis, George N.; Panagiotou, Nikolaos P.; Pyrgiotakis, Emmanouil G. |
| Abstract: | We combine machine learning algorithms (ML) with textual analysis techniques to forecast bank stock returns. Our textual features are derived from press releases of the Federal Open Market Committee (FOMC). We show that ML models produce more accurate out-of-sample predictions than OLS regressions, and that textual features can be more informative inputs than traditional financial variables. However, we achieve the highest predictive accuracy by training ML models on a combination of both financial variables and textual data. Importantly, portfolios constructed using the predictions of our best performing ML model consistently outperform their benchmarks. Our findings add to the scarce literature on bank return predictability and have important implications for investors. |
| Keywords: | Bank stock prediction; Trading strategies; Machine learning; Press conferences; Natural language processing; Banks |
| JEL: | C53 C88 G00 G11 G12 G14 G17 G21 |
| Date: | 2024–10 |
| URL: | https://d.repec.org/n?u=RePEc:pra:mprapa:122899 |
| By: | Ignacio Garr\'on; Andrey Ramos |
| Abstract: | Accurate tracking of anthropogenic carbon dioxide (CO2) emissions is crucial for shaping climate policies and meeting global decarbonization targets. However, energy consumption and emissions data are released annually and with substantial publication lags, hindering timely decision-making. This paper introduces a panel nowcasting framework to produce higher-frequency predictions of the state-level growth rate of per-capita energy consumption and CO2 emissions in the United States (U.S.). Our approach employs a panel mixed-data sampling (MIDAS) model to predict per-capita energy consumption growth, considering quarterly personal income, monthly electricity consumption, and a weekly economic conditions index as predictors. A bridge equation linking per-capita CO2 emissions growth with the nowcasts of energy consumption is estimated using panel quantile regression methods. A pseudo out-of-sample study (2009-2018), simulating the real-time data release calendar, confirms the improved accuracy of our nowcasts with respect to a historical benchmark. Our results suggest that by leveraging the availability of higher-frequency indicators, we not only enhance predictive accuracy for per-capita energy consumption growth but also provide more reliable estimates of the distribution of CO2 emissions growth. |
| Date: | 2025–01 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2501.03380 |