|
on Forecasting |
| By: | Hao Wu; David Levinson (TransportLab, School of Civil Engineering, University of Sydney) |
| Abstract: | Ensemble forecasting is a modeling approach that combines data sources, models of different types, with alternative assumptions, using distinct pattern recognition methods. The aim is to use all available information in predictions, without the limiting and arbitrary choices and dependencies resulting from a single statistical or machine learning approach or a single functional form, or results from a limited data source. Uncertainties are systematically accounted for. Outputs of ensemble models can be presented as a range of possibilities, to indicate the amount of uncertainty in modeling. We review methods and applications of ensemble models both within and outside of transport research. The review finds that ensemble forecasting generally improves forecast accuracy, robustness in many fields, particularly in weather forecasting where the method originated. We note that ensemble methods are highly siloed across different disciplines, and both the knowledge and application of ensemble forecasting are lacking in transport. In this paper we review and synthesize methods of ensemble forecasting with a unifying framework, categorizing ensemble methods into two broad and not mutually exclusive categories, namely combining models, and combining data; this framework further extends to ensembles of ensembles. We apply ensemble forecasting to transport related cases, which shows the potential of ensemble models in improving forecast accuracy and reliability. This paper sheds light on the apparatus of ensemble forecasting, which we hope contributes to the better understanding and wider adoption of ensemble models. |
| Keywords: | Ensemble forecasting, Combining models, Data fusion, Ensembles of ensembles |
| JEL: | R41 C93 |
| Date: | 2022 |
| URL: | http://d.repec.org/n?u=RePEc:nex:wpaper:ensembleapproachforecasting&r= |
| By: | Byron Botha; Rulof Burger; Kevin Kotze; Neil Rankin,; Daan Steenkamp |
| Abstract: | We investigate whether the use of machine learning techniques and big data can enhance the accuracy of inflation forecasts and our understanding of the drivers of South African inflation. We make use of a large dataset for the disaggregated prices of consumption goods and services to compare the forecasting performance of a suite of different statistical learning models to several traditional time series models. We find that the statistical learning models are able to compete with most benchmarks, but their relative performance is more impressive when the rate of inflation deviates from its steady state, as was the case during the recent COVID-19 lockdown, and where one makes use of a conditional forecasting function that allows for the use of future information relating to the evolution of the inflationary process. We find that the accuracy of the Reserve Bank’s near-term inflation forecasts compare favourably to those from the models considered, reflecting the inclusion of off-model information such as electricity tariff adjustments and within-month data. Lastly, we generate Shapley values to identify the most important contributors to future inflationary pressure and provide policymakers with information about the potential sources of future inflationary pressure. |
| Keywords: | Micro-data, Inflation, High dimensional regression, Penalised likelihood, Bayesian methods, Statistical learning |
| JEL: | C10 C11 C52 C55 E31 |
| Date: | 2022–02 |
| URL: | http://d.repec.org/n?u=RePEc:rza:wpaper:873&r= |
| By: | Kazeem Isah; Abdulkader Cassim Mahomedy; Elias Udeaja; Ojo Adelakun; Yusuf Yakubua |
| Abstract: | Motivated by the distinctive paradoxical nature of the Nigerian economy as the only OPEC oil-exporting economy that yet depends heavily on the importation of gasoline, we are compelled to re-examine the accuracy of the oil-based augmented Philips curve model in the predictability of inflation. Using quarterly data from 1970 to 2020, we investigate whether including the exchange rate into the oil price-based augmented Phillips curve improves the accuracy of forecasting inflation for the Nigerian economy. We rely on the outcomes of our preliminary analysis to account for the presence of endogeneity, persistence, and conditional heteroscedasticity in the predictability of inflation following the Westerlund & Narayan (2015) procedure. We find the extended variant of the oil price-based Phillips curve model that includes the exchange rate pass-through as most accurate for improving inflation forecasts in Nigeria. Given the robustness of our results from several models, we conclude that the exchange rate channel through which shocks to the oil price transmit into the economy is essential for forecasting inflation. |
| Keywords: | Nigeria, Inflation forecasts, Phillips curve, Oil price-exchange rate asymmetry |
| JEL: | E53 E31 E37 |
| Date: | 2022–02 |
| URL: | http://d.repec.org/n?u=RePEc:rza:wpaper:875&r= |
| By: | Stephen Bazen (AMSE - Aix-Marseille Sciences Economiques - EHESS - École des hautes études en sciences sociales - AMU - Aix Marseille Université - ECM - École Centrale de Marseille - CNRS - Centre National de la Recherche Scientifique); Jean Marie Cardebat (BSE - Bordeaux Sciences Economiques - UB - Université de Bordeaux - CNRS - Centre National de la Recherche Scientifique) |
| Abstract: | In an earlier article, we found that a univariate state space time series model estimated using the Kalman filter provided reasonably accurate monthly forecasts of generic Bordeaux red wine prices for the period up to 2016. We use the same model to forecast prices for the period 2017 to 2020, a period in which the market for this wine was subject to a number of demand and supply shocks. We find that the model's forecasts are poor from late 2018 through 2019 when the price collapsed from an all-time high and returned levels not seen since 2012. There is evidence of a structural break or regime change, and we explore the underlying reasons for this. The main one is the collapse in Chinese demand for Bordeaux wines which was brought about by a combination of shocks. |
| Date: | 2022–02 |
| URL: | http://d.repec.org/n?u=RePEc:hal:journl:hal-03603071&r= |
| By: | Mr. Jean-Francois Dauphin; Marzie Taheri Sanjani; Mrs. Nujin Suphaphiphat; Mr. Kamil Dybczak; Hanqi Zhang; Morgan Maneely; Yifei Wang |
| Abstract: | This paper describes recent work to strengthen nowcasting capacity at the IMF’s European department. It motivates and compiles datasets of standard and nontraditional variables, such as Google search and air quality. It applies standard dynamic factor models (DFMs) and several machine learning (ML) algorithms to nowcast GDP growth across a heterogenous group of European economies during normal and crisis times. Most of our methods significantly outperform the AR(1) benchmark model. Our DFMs tend to perform better during normal times while many of the ML methods we used performed strongly at identifying turning points. Our approach is easily applicable to other countries, subject to data availability. |
| Keywords: | Nowcasting, Factor Model, Machine Learning, Large Data Sets |
| Date: | 2022–03–11 |
| URL: | http://d.repec.org/n?u=RePEc:imf:imfwpa:2022/052&r= |
| By: | François Facchini (CES - Centre d'économie de la Sorbonne - UP1 - Université Paris 1 Panthéon-Sorbonne - CNRS - Centre National de la Recherche Scientifique) |
| Abstract: | This article is an update and extension of the electoral forecasting model of Lafay, Facchini and Auberger (2007) for the French presidential elections of 2022. Lafay and al. argued that the Socialist Party's rating was a good way to predict the vote split in the second round of elections between the left and the right. Socialist Pary's rating, nonetheless, cannot explain Emmanuel Macron's victory in the 2017 elections. This does not mean that party ratings are not a good predictor of the 2022 elections, if a number of adjustments are made. Based on party ratings the indicators proposed in this article argue that the scores in the first round of the April 2022 elections should be as follows: 30.5% for Emmanuel Macron, 22.7% for Valérie Pécresse (all the candidates of right wing), 18,7% for Marine Le Pen and 24.7% for the left and far left. The second round Macron - Pécresse is favorable to Emmanuel Macron, but depends fundamentally on the vote transfers between the left and the outgoing President. If the left abstains and Marine Le Pen's election rallies to the candidate of the right (LR), then Valérie Pécresse can win with a score of 51% against 49%. |
| Date: | 2022–03–23 |
| URL: | http://d.repec.org/n?u=RePEc:hal:cesptp:hal-03624729&r= |