Aggregate Forecasts Research Articles

Long-term probabilistic forecasts of activity mitigation in English hospitals: a national elicitation exercise providing an outside view based on judgements of experts in support of the New Hospital Programme

ObjectivesThe planning process for a new hospital relies on assumptions about future levels of demand. Typically, such assumptions are characterised by point estimates, the flaw-of-averages, base-rate neglect and overoptimism from...

The potential of publicly available weather forecasts for market operations in aggregated photovoltaic plants

Uncertainty in solar generation forecasts affects electricity market operations, forcing producers to bid conservatively and, consequently, reducing the amount of solar energy fed into the grid. This study analyses the potential of aggregated forecasts to reduce the errors of energy generation in medium-sized PV plants. To this objective, publicly available forecasts from the Global Forecasting System were processed with a basic neural network to generate hourly energy forecasts for individual PV plants of 1.1 MW and for the aggregated production of three of them (3.3 MW). The validation was carried out considering typical day-ahead and intraday market horizons, two zones with different atmospheric complexity, and discerning three types of days in terms of the received irradiation. Results show that the aggregation of forecasts would be beneficial as a general strategy in all cases, but especially in the zone of complex atmospheric dynamics and for the day-ahead horizons (36-42 h).

Flexible global forecast combinations

Forecast combination – the aggregation of individual forecasts from multiple experts or models – is a proven approach to economic forecasting. To date, research on economic forecasting has concentrated on local combination methods, which handle separate but related forecasting tasks in isolation. Yet, it has been known for over two decades in the machine learning community that global methods, which exploit task-relatedness, can improve on local methods that ignore it. Motivated by the possibility for improvement, this paper introduces a framework for globally combining forecasts while being flexible to the level of task-relatedness. Through our framework, we develop global versions of several existing forecast combinations. To evaluate the efficacy of these new global forecast combinations, we conduct extensive comparisons using synthetic and real data. Our real data comparisons, which involve forecasts of core economic indicators in the Eurozone, provide empirical evidence that the accuracy of global combinations of economic forecasts can surpass local combinations.

Predictive modeling of dining facility waste by material type across time and geography

The increasing rate of global solid waste generation is startling. This is exacerbating the challenge of decreasing solid waste generation to reduce disposal in landfills. U.S. Army installations offer a unique qualitative and quantitative dataset across a span of geographical locations. Modeling prediction capability of dining facility (DFAC) solid waste streams using data from 11 Army installations was investigated to demonstrate aggregate building and material type generation forecasting. Solid waste generation data were collected, quantified, and categorized for diversion potential (e.g., source reduction, reuse, recycling, composting, etc.) of materials currently landfilled. Over one week, samples from one day of DFAC operations were collected for each installation. Materials from the samples were manually separated into 22 categories, weighed, and recorded. Results identified many solid waste stream materials with diversion potential. Five material types were down selected to construct and validate the linear regression model. The material types down selection was based on available data robustness and applicability beyond military contexts. A linear regression model was constructed for five material and building type combinations to avoid multiplication factor errors of coefficients for each independent variable. Results were statistically significant (p-value ≤ 0.05) for four of five modeling combination predictions. These results demonstrate the unique capability of predicting solid waste generation for the four statistically significant model combinations. Each of the four statistically significant model combinations differed in adjusted R2 values, ranging from 0.988 to 0.996. This study provides five linear regression model combinations with predictive power that could reduce the labor- and cost-intensive process of characterizing waste streams while increasing data availability across the continental U.S. to focus targeted source reduction efforts for dining settings.

Forecast Aggregation and Predictive Value

Forecast Aggregation and Predictive Value

A Heuristic for Combining Correlated Experts When There Are Few Data

It is intuitive and theoretically sound to combine experts’ forecasts based on their proven skills, while accounting for correlation among their forecast submissions. Simpler combination methods, however, which assume independence of forecasts or equal skill, have been found to be empirically robust, in particular, in settings in which there are few historical data available for assessing experts’ skill. One explanation for the robust performance by simple methods is that empirical estimation of skill and of correlations introduces error, leading to worse aggregated forecasts than simpler alternatives. We offer a heuristic that accounts for skill and reduces estimation error by utilizing a common correlation factor. Our theoretical results present an optimal form for this common correlation, and we offer Bayesian estimators that can be used in practice. The common correlation heuristic is shown to outperform alternative combination methods on macroeconomic and experimental forecasting where there are limited historical data. This paper was accepted by Ilia Tsetlin, behavioral economics and decision analysis. Supplemental Material: The data file is available at https://doi.org/10.1287/mnsc.2021.02009 .

Numerical Analysis of Error From Sampling of Alternatives in Logit-Based Demand Forecasting Models with Massive Choice Sets

Use of the multinomial logit (MNL) functional form is widespread in transportation and land use demand modeling. As the state-of-the-art evolves away from aggregate forecasting and toward activity-based models and microsimulation, MNL models are being asked to accommodate increasingly large choice sets. Different strategies exist to address the challenge of estimating models with very large choice sets, but perhaps none is as commonly employed as sampling of alternatives. The effect of sampling of alternatives on parameter estimation has received considerable attention in the scientific literature. Yet comparatively little quantitative research exists that examines the issues that arise when the same sampling strategies are applied in the context of forecasting. In this study we conducted an analysis of the effect of sampling of alternatives in discrete choice models for disaggregate location choice forecasting. First, a novel measure of forecast error was defined and then used to quantify the extent of the problem as a function of sample rate, model error, and size of the universe of alternatives. Finally, we explored the potential for strategic sampling techniques to reduce the error observed under simple random sampling. In general, we found that the proportion of aggregate demand that was misallocated owing to sampling of alternatives was actually reduced as the size of the universe of alternatives increased. Additionally, simple random sampling was shown to outperform importance sampling in most scenarios. These findings suggest that error resulting from random sampling of alternatives is less of a problem for logit-based microsimulation models than previously assumed.

Forecasting with jury-based probabilistic argumentation

Probabilistic Argumentation naturally supports the integration of quantitative (probabilistic) reasoning and qualitative argumentation. Meanwhile, Jury-based Probabilistic Argumentation supports the combination of opinions by different reasoners. We show how Jury-based Probabilistic Argumentation (JPAA) and a form of Jury-based Probabilistic Assumption-based Argumentation (JPABA) can naturally support forecasting, whereby subjective probability estimates are combined to make predictions about future events. The form of JPABA we consider is an instance of JPAA and results from integrating Assumption-Based Argumentation (ABA) and probability spaces expressed by Bayesian networks. We show how JPAA and (the considered form of) JPABA can support forecasting by allowing different forecasters to determine the probability of arguments (and, in JPABA, sentences) with respect to their own probability spaces, while sharing arguments (and, in JPABA, their components). We show in turn how this supports the aggregation of individual forecasts into group forecasts.

Supplemental Material for Using Cross-Domain Expertise to Aggregate Forecasts When Within-Domain Expertise Is Unknown

Supplemental Material for Using Cross-Domain Expertise to Aggregate Forecasts When Within-Domain Expertise Is Unknown

Post-Pandemic Steel Production Scenarios for Poland Based on Forecasts of Annual Steel Production Volume

Abstract The paper presents the results of forecasts made for the volume of steel production in Poland based on actual data for the period from 2006 to 2021 with forecasting until 2026. The actual data used for the forecasts included annual steel production volumes in Poland (crude steel) in millions of tons. Basic adaptive methods were used to forecast the volume of steel production for the next five years. When selecting the methods, the course of the trend of the studied phenomenon was taken into account. In order to estimate the level of admissibility of the adopted forecasting methods, as well as to select the best forecasts, the errors of apparent forecasts (ex post) were calculated. Errors were calculated in the work: RMSE Root Mean Square Error being the square root of the mean square error of the ex-post forecasts yt for the period 2006-2021; ? as the mean value of the relative error of expired forecasts y*t (2006-2021) – this error informs about the part of the absolute error per unit of the real value of the variable yt. Optimization of the forecast values was based on the search for the minimum value of one of the above-mentioned errors, treated as an optimization criterion. In addition, the value of the point forecast (for 2022) obtained on the basis of the models used was compared with the steel production volume obtained for 3 quarters of 2022 in Poland with the forecast for the last quarter. Forecasting results obtained on the basis of the forecasting methods used, taking into account the permissible forecast errors, were considered as the basis for determining steel production scenarios for Poland until 2026. To determine the scenarios, forecast aggregation was used, and so the central forecasts were determined separately for decreasing trends and for increasing trends, based on the average values of the forecasts obtained for the period 2022-2026. The central forecasts were considered the baseline scenarios for steel production in Poland in 2022-2026 and the projected production volumes above the baseline forecasts with upward trends were considered an optimistic scenario, while the forecasted production volumes below the central scenario for downward trends were considered a pessimistic scenario for the Polish steel industry.

Dynamic information aggregation: Learning from the past

With dispersed information, how much can agents learn from past endogenous aggregates such as prices or output? In a rational-expectations equilibrium, if general equilibrium effects are strong enough, aggregates no longer perfectly reveal underlying fundamentals. In this confounding regime, the effects of informational frictions are persistent over time, and the aggregate outcome displays an initial under-reaction followed by a delayed over-reaction relative to its perfect-information counterpart. In a standard New Keynesian model, we show that endogenous information aggregation helps bring the model predictions on aggregate forecasts closer to the data.

A market for trading forecasts: A wagering mechanism

In many areas of industry and society, including energy, healthcare, and logistics, agents collect vast amounts of data that are deemed proprietary. These data owners extract predictive information of varying quality and relevance from data depending on quantity, inherent information content, and their own technical expertise. Aggregating these data and heterogeneous predictive skills, which are distributed in terms of ownership, can result in a higher collective value for a prediction task. In this paper, a platform for improving predictions via the implicit pooling of private information in return for possible remuneration is envisioned. Specifically, a wagering-based forecast elicitation market platform has been designed, in which a buyer intending to improve their forecasts posts a prediction task, and sellers respond to it with their forecast reports and wagers. This market delivers an aggregated forecast to the buyer (pre-event) and allocates a payoff to the sellers (post-event) for their contribution. A payoff mechanism is proposed and it is proven that it satisfies several desirable economic properties, including those specific to electronic platforms. Furthermore, the properties of the forecast aggregation operator and scoring rules are discussed in order to emphasize their effect on the sellers’ payoff. Finally, numerical examples are provided in order to illustrate the structure and properties of the proposed market platform.

Insights into the accuracy of social scientists' forecasts of societal change.

How well can social scientists predict societal change, and what processes underlie their predictions? To answer these questions, we ran two forecasting tournaments testing the accuracy of predictions of societal change in domains commonly studied in the social sciences: ideological preferences, political polarization, life satisfaction, sentiment on social media, and gender-career and racial bias. After we provided them with historical trend data on the relevant domain, social scientists submitted pre-registered monthly forecasts for a year (Tournament 1; N = 86 teams and 359 forecasts), with an opportunity to update forecasts on the basis of new data six months later (Tournament 2; N = 120 teams and 546 forecasts). Benchmarking forecasting accuracy revealed that social scientists' forecasts were on average no more accurate than those of simple statistical models (historical means, random walks or linear regressions) or the aggregate forecasts of a sample from the general public (N = 802). However, scientists were more accurate if they had scientific expertise in a prediction domain, were interdisciplinary, used simpler models and based predictions on prior data.

Forecasting tourism demand of tourist attractions during the COVID-19 pandemic

ABSTRACT Due to the COVID-19 outbreak, forecasting the tourism demand of tourist attractions is facing unprecedented difficulties given the lack of understanding about the pandemic impacts and the unavailability of post-pandemic data for generating forecasts. In this study, two strategies are proposed to improve forecasting performance and address the above difficulties. First, a novel COVID-19 impact indicator is built to reflect the impacts of the pandemic on tourism demand. Second, an effective forecast aggregation algorithm is developed to efficiently generate forecasts despite limited post-pandemic data availability. To validate the effectiveness of these strategies, an empirical study using real data from a tourist attraction is conducted, and results demonstrate that these strategies improve the overall forecast performance, including forecast accuracy and stability.

Disaggregating sales prediction: A gravitational approach

Whenever companies plan to enter new geographical areas, they need disaggregated sales in each location. To make such predictions, sales time series or final customers’ data in geographical disaggregation are necessary. However, for most companies, such datasets are unavailable or impractical. The manuscript has two main goals. One focal problem is how to disaggregate an aggregate sales prediction with no historical proportions. The other is how to improve spatial models using Point of Interest (POI) data. To solve these problems, we combine two literature streams — spatial marketing and sales forecasting — and propose a new hybrid probabilistic approach: Gravitational Sales Prediction (GSP). Our approach uses POI data to estimate area attraction, customer stocks, and flows to predict sales proportions. We later use these proportions to disaggregate an aggregate forecast. GSP is validated using sales data from two countries and more than ten economic segments. When compared to a strong benchmark that relies on past sales proportions, GSP exceeded expectations by achieving not only a similar performance to the benchmark but also outperforming it in some locations. It showed the most promising results in the middle level of aggregation. The result is a powerful and flexible approach that can be embedded in any decision support system.

From Proper Scoring Rules to Max-Min Optimal Forecast Aggregation

There are many ways to elicit honest probabilistic forecasts from experts. Once those forecasts are elicited, there are many ways to aggregate them into a single forecast. Should the choice of elicitation method inform the choice of aggregation method? In “From Proper Scoring Rules to Max-Min Optimal Forecast Aggregation,” Neyman and Roughgarden establish a connection between these two problems. To every elicitation method they associate the aggregation method that improves as much as possible upon the forecast of a randomly chosen expert, in the worst case. This association maps the two most widely used elicitation methods (Brier and logarithmic scoring) to the two most well-known aggregation methods (linear and logarithmic pooling). The authors show a number of interesting properties of this connection, including a natural axiomatization of aggregation methods obtained through the connection, as well as an algorithm for efficient no-regret learning of expert weights.

Adaptive individual residential load forecasting based on deep learning and dynamic mirror descent

With a growing penetration of renewable energy generation in the modern power networks, it has become highly challenging for network operators to balance electricity supply and demand. Residential load forecasting nowadays plays an increasingly important role in this aspect and facilitates various interactions between power networks and electricity users. While numerous research works have been proposed targeting at aggregate residential load forecasting, only a few efforts have been made towards individual residential load forecasting. The issue of volatility of individual residential load has never been addressed in forecasting. Thus, to fill this gap, this paper presents a deep learning method empowered with dynamic mirror descent for adaptive individual residential load forecasting. The proposed method is evaluated on a real-life Irish residential load dataset, and the experimental results show that it improves the prediction accuracy by 9.1% and 11.6% in the aspects of RMSE and MAE respectively in comparison with a benchmark method.

HYBRID PARTICLE SWARM OPTIMIZATION-GRAVITATIONAL SEARCH ALGORITHMS DEEP LEARNING NETWORKS TO SIMULTANEOUSLY PROJECT MULTIPLE CRUDE OIL PRICE

The conventional linear econometric and statistical models are not effective for forecasting the nonlinear and complex nature of crude oil prices. Computational intelligence techniques and hybrid modelling principles have been proposed to address this issue. Multiple forecasts can be combined using linear or nonlinear methods to create an aggregate forecast. Currently, there is no study on the optimization of deep learning with a hybrid of gravitational search algorithm (GSA) and particle swarm optimization (PSO) for forecasting crude oil price benchmarks, including WTI, Brent, and Dubai. Additionally, most studies have focused only on using West Texas Intermediate (WTI) crude oil spot prices as their benchmark. A Bidirectional Long Short-Term Memory with hybrid gravitational search algorithm (GSA) and particle swarm optimization (PSO) to forecast Crude Oil prices is proposed. The proposed model outperformed FNNPSOGSA, FNNGA, LSTM and BiLSTM models with root mean square errors (RMSE) of 0.0029, 0.0011, and 0.0029, respectively. The proposed model is suited for Crude Oil forecasting.

SIMULTANEOUS PROJECTION OF WORLD MULTIPLE CRUDE OIL PRICE BENCHMARKS VIA HYBRIDE OF PARTICLE SWARM OPTIMIZATION-GRAVITATIONAL SEARCH ALGORITHM REDESIGNED NEURAL NETWORKS

Conventional methods such as econometrical and statistical models are no longer feasible to handle the nonlinear, non-stationary, chaotic, volatile, and complex nature of crude oil prices due to their linearity nature; computational intelligence techniques were proposed to address these issues. More so, single modelling principle and traditional methods have not been effective enough in forecasting of crude oil prices, hybrid modelling principles have been employed to predict crude oil prices by researchers with better improvements. A group researchers opined that the use of hybrid models can also reduce the risk of choosing an inappropriate model because the use of a single model cannot always accurately forecast the extremely complex crude oil price time-series. A combination method can be applied to multiple forecasts and perform linear or nonlinear combinations of the forecasts, leading to an aggregate forecast. However, no research is found to have used the hybrid of gravitational search algorithm (GSA) and particle swarm optimization (PSO) to train artificial neural network (ANN) for simultaneous prediction of crude oil price benchmarks of WTI, Brent and Dubai. More so, most researchers have dwelled on only West Texas Intermediate (WIT) crude oil spot prices as their benchmark.

Observer-based power forecast of individual and aggregated offshore wind turbines

Abstract. Due to the increasing share of wind energy in the power system, minute-scale wind power forecasts have gained importance. Remote-sensing-based approaches have proven to be a promising alternative to statistical methods and thus need to be further developed towards an operational use, aiming to increase their forecast availability and skill. Therefore, the contribution of this paper is to extend lidar-based forecasts to a methodology for observer-based probabilistic power forecasts of individual wind turbines and aggregated wind farm power. To do so, lidar-based forecasts are combined with supervisory control and data acquisition (SCADA)-based forecasts that advect wind vectors derived from wind turbine operational data. After a calibration, forecasts of individual turbines are aggregated to a probabilistic power forecast of turbine subsets by means of a copula approach. We found that combining the lidar- and SCADA-based forecasts significantly improved both forecast skill and forecast availability of a 5 min ahead probabilistic power forecast at an offshore wind farm. Calibration further increased the forecast skill. Calibrated observer-based forecasts outperformed the benchmark persistence for unstable atmospheric conditions. The aggregation of probabilistic forecasts of turbine subsets revealed the potential of the copula approach. We discuss the skill, robustness and dependency on atmospheric conditions of the individual forecasts, the value of the observer-based forecast, its calibration and aggregation, and more generally the value of minute-scale power forecasts of offshore wind. In conclusion, combining different data sources to an observer-based forecast is beneficial in all regarded cases. For an operational use one should distinguish between and adapt to atmospheric stability.