Mean Forecast Error Research Articles

Time‐series‐based ensemble model output statistics for temperature forecasts postprocessing

AbstractThe uncertainty in numerical weather prediction models is nowadays quantified by the use of ensemble forecasts. Although these forecasts are continuously improved, they still suffer from systematic bias and dispersion errors. Statistical postprocessing methods, such as the ensemble model output statistics (EMOS), have been shown to substantially correct the forecasts. This work proposes an extension of EMOS in a time‐series framework. Besides taking account of seasonality and trend in the location and scale parameter of the predictive distribution, the autoregressive process in the mean forecast errors or the standardized forecast errors is considered. The models can be further extended by allowing generalized autoregressive conditional heteroscedasticity. Furthermore, it is outlined how to use these models for arbitrary forecast horizons. To illustrate the performance of the suggested EMOS models in time‐series fashion, we present a case study for the postprocessing of 2 m surface temperature forecasts using five different lead times and a set of observation stations in Germany. The results indicate that the time‐series EMOS extensions are able to significantly outperform the benchmark models EMOS and autoregressive EMOS (AR‐EMOS) in most of the lead time–station cases.

Are professional forecasters inattentive to public discussions about inflation? The case of Argentina

AbstractWe evaluate whether professional forecasters incorporate valuable information from public discussions on social media. The study covers the case of inflation in Argentina for the period 2016–2022. We find solid evidence consistent with inattention. A simple indicator of attention to inflation on social media is shown to anticipate professional forecast errors. A one standard deviation increment in the indicator is followed by a rise of 0.4% in mean forecast errors in the subsequent month and by a cumulative increment of 0.7% over the next 6 months. Furthermore, social media content anticipates significant revisions in forecasts that target multiple months ahead inflation and calendar year inflation. These findings are different from previously documented forms of inattention. Consistent results are verified by implementing out‐of‐sample forecasts and using content from an alternative social network. The study has implications for the use of professional forecasts in the context of policymaking and sheds new evidence on the nature of imperfect information in macroeconomics.

Prediction of Rainfall Trends Using Forecasting Approaches Based on Singular Spectrum Analysis

Advanced technologies such as the Internet of Things provide an integrated platform for weather focusing, including rainfall and flood prediction. Large rainfall data frequently contain noise, which can be difficult to analyze using a standard time series model due to violated assumptions. Singular spectrum analysis (SSA) is a model-free time series analysis method that is widely used. This study aims to predict the rainfall trends in the Special Region of Yogyakarta, Indonesia, using the Recurrent SSA (SSA-R) and Vector SSA (SSA-V). The SSA-R forecasts using the recurrent continuation directly with the linear recurrent formula, while the SSA-V is a modified recurrent method. This study used 50 years of monthly rainfall data (1970-2019) from 25 stations in the special region of Yogyakarta, Indonesia. The SSA steps for forecasting rainfall data include decomposition (embedding and singular value decomposition), reconstruction (grouping and diagonal averaging), and evaluating the SSA model using w-correlation (if w-correlation is close to zero, returning to the decomposition stage; otherwise, continue the process), forecasting, evaluating the forecast results using root mean square error (RMSE), mean absolute error, r, and mean forecast error, and finally selecting the best model (either the SSA-R or SSA-V model). The results showed that the SSA-R performed better than SSA-V due to the smallest RMSE in the dry, rainy, and inter-monsoon seasons. The SSA-R model’s forecast results revealed faint, constant patterns for the dry, and rainy seasons and an increasing pattern for the inter-monsoon season. The novelty of this study is to compare the performance of the SSA-R and SSA-V models in the large rainfall data in the special region of Yogyakarta, Indonesia.

Performance of short-terms prediction methods of vertical total electron content using nonlinear autoregressive neuronal network and stochastic autoregressive model

In this contribution the performance of global short-term predictions methods of vertical total electron content (vTEC) is analyzed during high solar activity. Two kind of predicted global vTEC maps value every 1 h, one-day-ahead, are used. They are C1PG, produced by the Center for Orbit Determination in Europe (CODE), based on the extrapolation of Spherical Harmonic coefficient using Least-squares collocation and the M1PG, proposed in this work, based on multi-step Nonlinear Autoregressive Neural Network (NAR-NN).Global vTEC maps from CODE (CODG) along the year 2015, are used as reference data. The results are obtained for quiet and disturbed conditions, based on geomagnetic and ionospheric planetary indexes.The performance of the forecasting approach is extensively tested under different geospatial conditions. The testing results are very similar in terms of RMSE, as it has been found to range between 1.7 and 7 TECu. RMSE depend on the latitude sectors and geomagnetic conditions, in terms of Mean Forecast Error (MFE) the C1PG shows a clear systematic behavior being negative at Southern Hemisphere and positive at Northern Hemisphere. According to the Mean Absolute Percentage Error (MAPE) values, the relative behavior of vTEC prediction is better for M1PG than for C1PG specially during quiet days and at mid-high latitudes. In general, both models are less accuracy in the equatorial ionization anomaly region and the Southern Hemisphere.Other important contribution of this manuscript is the definition of a planetary ionospheric disturbance index, Wd, based on W-index. This parameter is useful to more complete definition in quiet and disturbed days selection, and it is shown that the dependence of the RMSE according to the latitudinal bands, it is strongly related with the respective value of Wd.

Operating Status Prediction Model at EV Charging Stations With Fusing Spatiotemporal Graph Convolutional Network

This article proposes the operating status prediction model at electric vehicle (EV) charging stations based on the spatiotemporal graph convolutional network (SGCN). The SGCN combines graph convolutional network (GCN) and the gated recurrent unit (GRU), alleviating the queuing time at charging stations due to the lack of information for EV users. First, an urban charging station-traffic flow model is established to portray the interrelationship between charging stations and traffic. Second, a multistep prediction model based on SGCN for operating status at charging stations is proposed to forecast the occupancy of charging stations over the next tens of minutes. The comparison case study with the forecast and actual data reveals that the mean forecast error is around 19.21% when estimating 18 min ahead. Incrementing errors are subtle even after adding random noise to the original data. Finally, the model is applied to charging guidance decisions. Our model can reduce the number of EV queuing by 60% during high charging demand. It also shortens the average charging waiting time by 4 min.

Detection of reinforcement of multi-bead and multi-layer weld in additive manufacturing based on on-line visual information of weld pool

In the multi-bead and multi-layer arc additive manufacturing process, the information of cladding reinforcement reflects the welding quality to a certain extent, so it is of great significance to monitor the reinforcement of cladding layers in real time. In this paper, a point cloud density search method is used to segment the point cloud of a single weld bead in the multi bead weld seam, and then the reinforcement of each cladding bead of multi-bead and multi-layer weld is extracted separately when the bottom plate is deformed due to high temperature, and a residual-based prediction model is constructed for quantitative forecasting of the transient reinforcement before solidification of block cladding layer in real time. The following work is completed to prove the accuracy and effectiveness of the proposed model, two different strategies are used to predict the reinforcement of multi-bead and multi-layer welds. Through the experiment, we can see the mean forecast error of the reinforcement of multi-bead and multi-layer welds is less than 0.3 mm, while the time for the model to dealing with the molten pool image is 18 ms, and the optimal strategy can make the average error better than 0.15 mm, which proves that the model constructed in this paper has great generalization performance and realizes the real-time and high-precision prediction of cladding reinforcement in the case of small deformation. The study of this paper supplies a necessary basis for the online monitoring and control of morphological defects in the process of weld processing.

Filter Likelihood as an Observation-Based Verification Metric in Ensemble Forecasting

In numerical weather prediction (NWP), ensemble forecasting aims to quantify the flow-dependent forecast uncertainty. The focus here is on observation-based verification of the reliability of ensemble forecasting systems. In particular, at short forecast lead times, forecast errors tend to be relatively small compared to observation errors and hence it is very important that the verification metric also accounts for observational uncertainties. This paper studies the so-called <em>filter likelihood score</em> which is deep-rooted in Bayesian estimation theory and fits naturally to the filtering setup of NWP. The filter likelihood score considers observation errors, ensemble mean skill, and ensemble spread in one metric. Importantly, it can be made multivariate and effortlessly expanded to simultaneous verification against all observation types through the observation operators contained in the parental data assimilation scheme. Here observations from the global radiosonde network and satellites (AMSU-A channel 5) are included in the verification of OpenIFS-based ensemble forecasts using different types of initial state perturbations. Our results show that the filter likelihood score is sensitive to the ensemble prediction system quality and compares consistently with other verification metrics such as the relationships between ensemble spread and ensemble mean forecast error, and Dawid-Sebastiani score. Our conclusion is that the filter likelihood score provides a very well-behaving verification metric, that can be made truly multivariate by including covariances, for ensemble prediction systems with a strong foundation in estimation theory.

Uniform predictive inference for factor models with instrumental and idiosyncratic betas

This paper investigates the impact of allowing for characteristic-based time-varying factor betas on the diffusion-index type forecasts. The factor beta consists of two distinct components: the “instrumental beta” is a function of some observable characteristics, while the “idiosyncratic beta” captures more volatile residual movements. To estimate these characteristic-based time-varying betas and the corresponding factors, we apply the projected principal component analysis (P-PCA) method on high-frequency returns data. The primary advantage of this method is that it refines the estimators of latent factors, which shall be used in the forecasting models. We show that various leading components of the conditional mean forecast error are all asymptotically normal and pairwise independent. Extensive simulation studies show the good finite-sample properties of the P-PCA estimators and demonstrate the advantage of the P-PCA method relative to the classic PCA method in forecasting. In our empirical experiments of volatility prediction, we find that the factor-augmented model associated with the P-PCA method is more parsimonious and achieves better performance for a wide variety of target assets. We also find evidence on different levels of variation over time in the idiosyncratic beta, which necessitates our uniform predictive inference procedure.

Option price predictability, splines, and expanded rationality

The current practice of option price forecast relies on the outputs of various option pricing models. The expected value of the current option price is widely regarded as the best forecast for the future price, assuming the option prices evolve with a Brownian motion. However, volatility clustering, transaction illiquidity, and demand-supply imbalance drive the future option prices off the modeled price targets. Therefore, we suggest using the spline method to forecast option prices directly. The focus is the accuracy of the forecasted asset price in the next period, rather than if the pricing models correctly produce the current price. We use fifteen years of daily SPY American option contract prices to examine the spline model forecast accuracy. Among the 476,882 forecasts produced, the mean forecasting error size is $3.66 × 10-3, with a standard deviation of 1.33 and a median error of $5.54 × 10-17. The forecast accuracy is stable across contracts with different terms and moneyness. The spline forecast model incorporates the illiquidity issue and avoids the vital pitfalls in the current leading option pricing techniques.

Cross-country uncertainty spillovers: Evidence from international survey data

Using a large international survey of professional forecasters, we construct measures of economic uncertainty surrounding output growth, inflation, the interest rate, exchange rate and current account. We then analyze uncertainty spillovers across major advanced and emerging economies using large multi-country Bayesian Panel VARs. We consider how our results change if our uncertainty measures reflect: disagreement among forecasters (idiosyncratic uncertainty); the variance of their mean forecast errors (common uncertainty); or both types of uncertainty. We show that the US is an important but not dominant source of uncertainty, affecting other economies through interest rate and exchange rate uncertainty. This reflects the major role played by US monetary policy and the dollar in the global financial system. Crucially, though, the Eurozone followed by the UK and China are also important sources of uncertainty. We also find that, on average, foreign interest rate and exchange rate uncertainty are more important than foreign output growth uncertainty. While spillovers in idiosyncratic uncertainty are more frequently observed, failing to account for common uncertainty can lead us to overestimate the role played by smaller economies.

An Initial Field Intelligent Correcting Algorithm for Numerical Forecasting Based on Artificial Neural Networks under the Conditions of Limited Observations: Part I—Focusing on Ocean Temperature

For the numerical forecasting of ocean temperature, the effective fusion of observations and the initial field under the conditions of limited observations has always been a significant problem. Traditional data assimilation methods cannot make full use of limited observations to correct the initial field. In order to obtain an optimal initial field with limited observations, this study proposed an intelligent correcting (IC) algorithm based on artificial neural networks (ANNs). The IC algorithm can fully mine the correlation laws between the grid points using historical data, and this process essentially replaces the estimation of background error covariance in traditional data assimilation methods. Experimental results show that the IC algorithm can lead to superior forecasting accuracy, with a lower root mean square error (around 0.7 °C) and higher coefficient of determination (0.9934) relative to the optimal interpolation method. Through the IC algorithm, the largest reduction in mean forecasting error can reach around −0.5 °C and the maximum percentage decline in mean forecasting error can reach 30% compared with the original numerical forecasting results. Therefore, the experiments validate that the IC algorithm can effectively correct the initial field under the conditions of limited observations.

Function fitting for modeling seasonal normalized difference vegetation index time series and early forecasting of soybean yield

Forecasting crop yields based on remote sensing data is one of the most important tasks in agriculture. Soybean is the main crop in the Russian Far East. It is desirable to forecast soybean yield as early as possible while maintaining high accuracy. This study aimed to investigate seasonal time series of the normalized difference vegetation index (NDVI) to achieve early forecasting of soybean yield. This research used data from the Moderate Resolution Image Spectroradiometer (MODIS), an arable-land mask obtained from the VEGA-Science web service, and soybean yield data for 2008–2017 for the Jewish Autonomous Region (JAR) districts. Four approximating functions were fitted to model the NDVI time series: Gaussian, double logistic (DL), and quadratic and cubic polynomials. In the period from calendar weeks 22–42 (end of May to mid-October), averaged over two districts, the model using the DL function showed the highest accuracy (mean absolute percentage error –4.0%, root mean square error (RMSE) –0.029, P < 0.01). The yield forecast accuracy of prediction in the period of weeks 25–30 in JAR municipalities using the parameters of the Gaussian function was higher (P < 0.05) than that using the other functions. The mean forecast error for the Gaussian function was 14.9% in week 25 (RMSE was 0.21 t ha−1) and 5.1%−12.9% in weeks 26–30 (RMSE varied from 0.06 to 0.15 t ha−1) according to the 2013–2017 data. In weeks 31–32, the error was 5.0%−5.4% (RMSE was 0.07 t ha−1) using the Gaussian parameters and 7.4%−7.7% (RMSE was 0.09–0.11 t ha−1) for the DL function. When the method was applied to municipal districts of other soy-producing regions of the Russian Far East. RMSE was 0.14–0.32 t ha−1 in weeks 25–26 and did not exceed 0.20 t ha−1 in subsequent weeks.

Procurement Practice of Program Drugs and Its Challenges at the Ethiopian Pharmaceuticals Supply Agency: A Mixed Methods Study.

BackgroundEffective drug procurement guarantees the sustainable supply of products for health and eliminates excessive costs. However, there is limited information on the area of pharmaceutical procurement practice in Ethiopia. Thus, this study aimed at assessing the procurement practices of program drugs and its challenges at the Ethiopian Pharmaceuticals Supply Agency.MethodsA cross-sectional study accompanied by qualitative assessment was conducted between February 21 and April 20/2020 to examine the procurement practice of the Ethiopian pharmaceutical supply agency. The quantitative data were gathered by reviewing documents and electronic records. Mean forecast error, price paid to international price reference, number of emergency orders placed, and lead time variability were the measurements used to assess the procurement practice. A statistical package for the social sciences version 23 was used to analyze the data. The results were then summarized using tables and texts. The qualitative data were collated through face-to-face in-depth interviews to explore the challenges behind the procurement practice. And the data were analyzed manually using the thematic analysis technique.ResultsThe agency had its own procurement list which defines the items to be procured. The overall mean forecast error in the 2018/19 budget year was 27.8%. Of the 70 program commodities included in the study, 52 (74.3%) items had a mean price less than the international price reference. Three of the 14 orders (21.4%) placed in the aforementioned year were emergency purchases made through direct procurement. The mean lead time for the suppliers of the agency was 137.3 days. Poor data quality from service delivery points, staff capacity constraints, communication problems, and policy issues became the major challenges to implement an effective procurement system in the agency.ConclusionThe procurement practice at the agency has strong side. However, it was not without weaknesses. Using a procurement list is a worthwhile practice. Despite this, much remains to improve lead times and forecasting accuracy.

Bitcoin daily close price prediction using optimized grid search method

Abstract Cryptocurrencies are digital assets that can be stored and transferred electronically. Bitcoin (BTC) is one of the most popular cryptocurrencies that has attracted many attentions. The BTC price is considered as a high volatility time series with non-stationary and non-linear behavior. Therefore, the BTC price forecasting is a new, challenging, and open problem. In this research, we aim the predicting price using machine learning and statistical techniques. We deploy several robust approaches such as the Box-Jenkins, Autoregression (AR), Moving Average (MA), ARIMA, Autocorrelation Function (ACF), Partial Autocorrelation Function (PACF), and Grid Search algorithms to predict BTC price. To evaluate the performance of the proposed model, Forecast Error (FE), Mean Forecast Error (MFE), Mean Absolute Error (MAE), Mean Squared Error (MSE), as well as Root Mean Squared Error (RMSE), are considered in our study.

Short-term fire forecast based on air state gain recurrence and zero-order brown model

Possibilities of parameterization of the zero-order Brown model for indoor air forecasting based on the current measure of air state gain recurrence are considered. The key to the zero-order parametric Brown forecasting model is the selection of the smoothing parameter, which characterizes forecast adaptability to the current air state gain recurrence measure. It is shown that for effective short-term indoor fire forecast, the Brown model parameter must be selected from the out-of-limit set defined by 1 and 2. The out-of-limit set for the Brown model parameter is an area of effective fire forecasting based on the measure of current indoor air state gain recurrence. Errors of fire forecast based on the parameterized zero-order Brown model in the case of the classical and out-of-limit sets of the model parameters are investigated using the example of ignition of various materials in a laboratory chamber. As quantitative indicators of forecast quality, the absolute and mean forecast errors exponentially smoothed with a parameter of 0.4 are investigated. It was found that for alcohol, the smoothed absolute and mean forecast errors for the classical smoothing parameter in the no-ignition interval do not exceed 20 %. At the same time, for the out-of-limit case, the indicated forecast errors are, on average, an order of magnitude smaller. Similar ratios for forecast errors remain in paper, wood and textile ignition. However, for the transition zone corresponding to the time of material ignition, a sharp decrease in the current measure of chamber air state gain recurrence is observed. It was found that for this zone, the smoothed absolute forecast error for alcohol is about 2 % if the model parameter is selected from the classical set. If the model parameter is selected from the out-of-limit set, the forecast error is about 0.2 %. The results generally demonstrate significant advantages of using the zero-order Brown parametric model with out-of-limit model parameters for indoor fire forecasting

Simulation and optimal control of heating and cooling systems: A case study of a commercial building

This paper proposes a novel energy conservation measure that optimizes the planning of heating and cooling systems for tertiary sector buildings. It consists of a model-based predictive control approach that employs a grey-box model built from the building and weather data that predicts the building heat load and indoor temperature. Different from classical optimization approaches where the discretized differential algebraic equations are integrated into the optimization formulation, our model is calibrated using black-box multiobjective optimization, which allows for decoupling the predictive model from the optimization problem, thus having more flexibility and reducing the total computational time. Moreover, rather than requiring the angle of solar radiation, solar orientation and solar masks to calculate the radiation data, our approach requires only a simple model of the solar irradiance. The calibrated model is then used by heating and cooling optimization strategies that aim at reducing the energy consumption of the building in the next day while satisfying the indoor thermal constraints. The proposed approach was applied in a case study of a commercial building during heating and cooling seasons and the results show that it was able to yield up to 12% of energy savings while having a mean power forecast error of 8%.

Performance Evaluation of Different Membership Function in Fuzzy Logic Based Short-Term Load Forecasting

A mismatch between utility-scale electricity generation and demand often results in resources and energy wastage that needed to be minimized. Therefore, the utility company needs to be able to accurately forecast load demand as a guide for the planned generation. Short-term load forecast assists the utility company in projecting the future energy demand. The predicted load demand is used to plan ahead for the power to be generated, transmitted, and distributed and which is crucial to power system reliability and economics. Recently, various methods from statistical, artificial intelligence, and hybrid methods have been widely used for load forecasts with each having their merits and drawbacks. This paper investigates the application of the fuzzy logic technique for short-term load forecast of a day ahead load. The developed fuzzy logic model used time, temperature, and historical load data to forecast 24 hours load demand. The fuzzy models were based on both the trapezoidal and triangular membership function (MF) to investigate their accuracy and effectiveness for the load forecast. The obtained low Mean Absolute Percentage Error (MAPE), Mean Forecast Error (MFE), and Mean Absolute Deviation (MAD) values from the forecasted load results showed that both models are suitable for short-term load forecasting, however the trapezoidal MF showed better performance than the triangular MF.

Comparing SANN and SARIMA for forecasting frequency of monthly rainfall in Umuahia

This paper adopted the use of artificial neural network (ANN) which is also a nonlinear model in contributing to the debate of comparing different methods of forecasting frequency of rainfall in Umuahia, Abia State of Nigeria. The choice of ANN is informed by articles in literature which have shown that Neural Network models outperform some traditional statistical models in modelling meteorological time series. The data used for the research was obtained from the National Root Crop Research Institute, Umudike, Abia State and it spanned from 2006 to 2016. The forecasting performance of seasonal autoregressive integrated moving average (SARIMA) model and that of seasonal artificial neural network (SANN) were compared with four forecast performance measures:- Forecast Error (FE), Mean Forecast Error (MFE), Mean Squared Error (MSE) and Root Mean Squared Error (RMSE). Results from the study showed that the SARIMA had lower error indicators in forecast performance and thus was adjudged better than SANN in forecasting frequency of rainfall in Umuahia, Abia State. A t-test for significant difference showed that there is no statistical significant difference between both forecast values. Hence, the authors conclude that the two methods can be successfully used as substitutes. Farmers, Construction Companies, Government agencies working in Umuahia are therefore advised to choose SARIMA modelling of monthly rainfall frequency over that of SANN. This study also confirms that ANNs are not utterly better than the traditional statistical models as seen in other works in literature.

Understanding Error Distributions of Hurricane Intensity Forecasts during Rapid Intensity Changes

AbstractThe characteristics of official National Hurricane Center (NHC) intensity forecast errors are examined for the North Atlantic and east Pacific basins from 1989 to 2018. It is shown how rapid intensification (RI) and rapid weakening (RW) influence yearly NHC forecast errors for forecasts between 12 and 48 h in length. In addition to being the tail of the intensity change distribution, RI and RW are at the tails of the forecast error distribution. Yearly mean absolute forecast errors are positively correlated with the yearly number of RI/RW occurrences and explain roughly 20% of the variance in the Atlantic and 30% in the east Pacific. The higher occurrence of RI events in the east Pacific contributes to larger intensity forecast errors overall but also a better probability of detection and success ratio. Statistically significant improvements to 24-h RI forecast biases have been made in the east Pacific and to 24-h RW biases in the Atlantic. Over-ocean 24-h RW events cause larger mean errors in the east Pacific that have not improved with time. Environmental predictors from the Statistical Hurricane Intensity Prediction Scheme (SHIPS) are used to diagnose what conditions lead to the largest RI and RW forecast errors on average. The forecast error distributions widen for both RI and RW when tropical systems experience low vertical wind shear, warm sea surface temperature, and moderate low-level relative humidity. Consistent with existing literature, the forecast error distributions suggest that improvements to our observational capabilities, understanding, and prediction of inner-core processes is paramount to both RI and RW prediction.

Comparison of daily rainfall forecasting using multilayer perceptron neural network model

Rainfall is important in predicting weather forecast particularly to the agriculture sector and also in environment which gives great contribution towards the economy of the nation. Thus, it is important for the hydrologists to forecast daily rainfall in order to help the other people in the agriculture sector to proceed with their harvesting schedules accordingly and to make sure the results of their crops would be satisfying. This study is set to forecast the daily rainfall future value using ARIMA model and Artificial Neural Network (ANN) model. Both method is evaluated by using Mean Absolute Error (MAE), Mean Forecast Error (MFE), Root Mean Squared Error (RMSE) and coefficient of determination (R ). The results showed that ANN model has outperformed results than ARIMA model. The results also showed ANN has under-forecast the daily rainfall data by 2.21% compare to ARIMA with over-forecast of -3.34%. From this study, it shows that the ANN (6,4,1) model produces better results of MAE (8.4208), MFE (2.2188), RMSE (34.6740) and R (0.9432) compared to ARIMA model. This has proved that ANN model has outperformed ARIMA model in predicting daily rainfall values.