|
on Forecasting |
| By: | Thomas Gaertner; Christoph Lippert; Stefan Konigorski |
| Abstract: | In response to the growing demand for accurate demand forecasts, this research proposes a generalized automated sales forecasting pipeline tailored for small- to medium-sized enterprises (SMEs). Unlike large corporations with dedicated data scientists for sales forecasting, SMEs often lack such resources. To address this, we developed a comprehensive forecasting pipeline that automates time series sales forecasting, encompassing data preparation, model training, and selection based on validation results. The development included two main components: model preselection and the forecasting pipeline. In the first phase, state-of-the-art methods were evaluated on a showcase dataset, leading to the selection of ARIMA, SARIMAX, Holt-Winters Exponential Smoothing, Regression Tree, Dilated Convolutional Neural Networks, and Generalized Additive Models. An ensemble prediction of these models was also included. Long-Short-Term Memory (LSTM) networks were excluded due to suboptimal prediction accuracy, and Facebook Prophet was omitted for compatibility reasons. In the second phase, the proposed forecasting pipeline was tested with SMEs in the food and electric industries, revealing variable model performance across different companies. While one project-based company derived no benefit, others achieved superior forecasts compared to naive estimators. Our findings suggest that no single model is universally superior. Instead, a diverse set of models, when integrated within an automated validation framework, can significantly enhance forecasting accuracy for SMEs. These results emphasize the importance of model diversity and automated validation in addressing the unique needs of each business. This research contributes to the field by providing SMEs access to state-of-the-art sales forecasting tools, enabling data-driven decision-making and improving operational efficiency. |
| Date: | 2024–12 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2412.20420 |
| By: | Faria, Gonçalo; Verona, Fabio |
| Abstract: | We introduce a frequency-domain forecast combination method that leverages time- and frequencydependent predictability to enhance forecast accuracy. By decomposing both the target variables (equity premium and real GDP growth) and predictor variables into distinct frequency components, this method aligns forecasts with frequency-specific predictive relationships. This approach yields significantly higher accuracy than traditional time-domain methods, as evidenced by both statistical and economic out-of-sample metrics. Gains are particularly pronounced during recessions, where excluding low-frequency components further enhances forecast precision. Overall, these findings highlight the value of frequency-domain forecasting in capturing complex, time-varying patterns across varied macro-financial contexts. |
| Keywords: | forecast combination, frequency domain, equity premium, GDP growth, Haar filter |
| JEL: | C58 G11 G17 |
| Date: | 2024 |
| URL: | https://d.repec.org/n?u=RePEc:zbw:bofrdp:307140 |
| By: | Verona, Fabio |
| Abstract: | This paper addresses the challenge of inflation forecasting by adopting a thick modeling approach that integrates forecasts from time- and frequency-domain models. Frequency-domain models excel at capturing long-term trends while also accounting for short-term fluctuations. Combining these models with traditional approaches leverages their complementary strengths, resulting in forecasts that consistently outperform individual methods, especially during periods of heightened inflation volatility. By pooling insights from diverse modeling frameworks, this study provides a robust and effective strategy for improving inflation forecasts across different horizons. |
| Keywords: | inflation forecasting, forecast combination, wavelets, Haar filter, time-varying parameters, Phillips curve |
| JEL: | C32 C53 E31 E37 E43 E44 |
| Date: | 2025 |
| URL: | https://d.repec.org/n?u=RePEc:zbw:bofrdp:308098 |
| By: | Francesco Audrino (University of St. Gallen; Swiss Finance Institute); Jonathan Chassot (University of St. Gallen) |
| Abstract: | We investigate the predictive abilities of the heterogeneous autoregressive (HAR) model compared to machine learning (ML) techniques across an unprecedented dataset of 1, 445 stocks. Our analysis focuses on the role of fitting schemes, particularly the training window and re-estimation frequency, in determining the HAR model's performance. Despite extensive hyperparameter tuning, ML models fail to surpass the linear benchmark set by HAR when utilizing a refined fitting approach for the latter. Moreover, the simplicity of HAR allows for an interpretable model with drastically lower computational costs. We assess performance using QLIKE, MSE, and realized utility metrics, finding that HAR consistently outperforms its ML counterparts when both rely solely on realized volatility and VIX as predictors. Our results underscore the importance of a correctly specified fitting scheme. They suggest that properly fitted HAR models provide superior forecasting accuracy, establishing robust guidelines for their practical application and use as a benchmark. This study not only reaffirms the efficacy of the HAR model but also provides a critical perspective on the practical limitations of ML approaches in realized volatility forecasting. |
| Keywords: | Forecasting practice, HAR, Machine learning, Realized volatility, Volatility forecasting |
| Date: | 2024–11 |
| URL: | https://d.repec.org/n?u=RePEc:chf:rpseri:rp2470 |
| By: | Kamil {\L}. Szyd{\l}owski; Jaros{\l}aw A. Chudziak |
| Abstract: | This paper investigates the application of Transformer-based neural networks to stock price forecasting, with a special focus on the intersection of machine learning techniques and financial market analysis. The evolution of Transformer models, from their inception to their adaptation for time series analysis in financial contexts, is reviewed and discussed. Central to our study is the exploration of the Hidformer model, which is currently recognized for its promising performance in time series prediction. The primary aim of this paper is to determine whether Hidformer will also prove itself in the task of stock price prediction. This slightly modified model serves as the framework for our experiments, integrating the principles of technical analysis with advanced machine learning concepts to enhance stock price prediction accuracy. We conduct an evaluation of the Hidformer model's performance, using a set of criteria to determine its efficacy. Our findings offer additional insights into the practical application of Transformer architectures in financial time series forecasting, highlighting their potential to improve algorithmic trading strategies, including human decision making. |
| Date: | 2024–12 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2412.19932 |
| By: | Mohammad R. Jahan-Parvar; Charles Knipp; Pawel J. Szerszen |
| Abstract: | We propose a generalized multivariate unobserved components model to decompose macroeconomic data into trend and cyclical components. We then forecast the series using Bayesian methods. We document that a fully Bayesian estimation, that accounts for state and parameter uncertainty, consistently dominates out-of-sample forecasts produced by alternative multivariate and univariate models. In addition, allowing for stochastic volatility components in variables improves forecasts. To address data limitations, we exploit cross-sectional information, use the commonalities across variables, and account for both parameter and state uncertainty. Finally, we find that an optimally pooled univariate model outperforms individual univariate specifications, andperforms generally closer to the benchmark model. |
| Keywords: | Bayesian estimation; Maximum likelihood estimation; Online forecasting; Out-of-sample forecasting; Parameter uncertainty; Sequential Monte Carlo methods; Trend-cycle decomposition |
| JEL: | C11 C22 C32 C53 |
| Date: | 2024–12–30 |
| URL: | https://d.repec.org/n?u=RePEc:fip:fedgfe:2024-100 |
| By: | Linzenich, Jan; Meunier, Baptiste |
| Abstract: | We provide a versatile nowcasting toolbox that supports three model classes (dynamic factor models, large Bayesian VAR, bridge equations) and offers methods to manage data selection and adjust for Covid-19 observations. The toolbox aims at simplifying two key tasks: creating new nowcasting models and improving the policy analysis. For model creation, the toolbox automatizes testing input variables, assessing model accuracy, and checking robustness to the Covid period. The toolbox is organized along a structured three-step approach: variable pre-selection, model selection, and Covid robustness. Non-specialists can easily follow these steps to develop high-performing models, while experts can leverage the automated tests and analyses. For regular policy use, the toolbox generates a large range of outputs to aid conjunctural analysis like news decomposition, confidence bands, alternative forecasts, and heatmaps. These multiple outputs aim at opening the "black box" often associated with nowcasts and at gauging the reliability of real-time predictions. We showcase the toolbox features to create a nowcasting model for global GDP growth. Overall, the toolbox aims at facilitating creation, evaluation, and deployment of nowcasting models. Code and templates are available on GitHub: https://github.com/baptiste-meunier/Nowcasting_toolbox. JEL Classification: C22, C51, C52, C53, C55 |
| Keywords: | Bayesian VAR, bridge equation, dynamic factor model, forecasting, large dataset |
| Date: | 2024–12 |
| URL: | https://d.repec.org/n?u=RePEc:ecb:ecbwps:20243004 |
| By: | Paul, Joseph R.; Schaffer, Mark E. |
| Abstract: | This paper introduces conformal inference, a powerful and flexible framework for constructing prediction intervals with guaranteed coverage in finite samples. Unlike conventional methods, conformal inference makes no assumptions about the underlying data distribution other than exchangeability. The paper begins with some simple examples of full and split conformal prediction that highlight the key assumption of exchangeability. We then provide more formal treatments of full and split conformal prediction along with extensions of the basic framework, including the Jackknife+ and CV+ algorithms, both of which offer a better balance between computational and statistical efficiency compared to full and split conformal prediction. The paper then discusses the limitations to achieving exact conditional coverage and several methods that aim to improve conditional coverage in practice. The final section briefly discusses areas of current research the software options for implementing conformal methods. |
| Keywords: | conformal inference, conformal prediction, distribution-free inference, machine learning |
| JEL: | C12 C14 C53 |
| Date: | 2024 |
| URL: | https://d.repec.org/n?u=RePEc:zbw:hwuaef:308058 |