Forecasting the NN5 time series with hybrid models

We propose a simple way of predicting time series with recurring seasonal periods. Missing values of the time series are estimated and interpolated in a preprocessing step. We combine several forecasting methods by taking the weighted mean of forecasts that were generated with time-domain models which were validated on left-out parts of the time series. The hybrid model is a combination of a neural network ensemble, an ensemble of nearest trajectory models and a model for the 7-day cycle. We apply this approach to the NN5 time series competition data set.     

Performance evaluation using data envelopment analysis in the presence of time lags

Data Envelopment Analysis (DEA) is a methodology that computes efficiency values for decision making units (DMU) in a given period by comparing the outputs with the inputs. In many applications, inputs and outputs of DMUs are monitored over time. There might be a time lag between the consumption of inputs and the production of outputs. We develop an approach that aims to capture the time lag between the outputs and the inputs in assigning the efficiency values to DMUs. We propose using weight restrictions in conjunction with the model. Our computational results on randomly generated problems demonstrate that the developed approach works well under a large variety of experimental conditions. We also apply our approach on a real data set to evaluate research institutions.     

Neural networks for GEFCom2017 probabilistic load forecasting

This report describes the forecasting model which was developed by team “4C” for the global energy forecasting competition 2017 (GEFCom2017), with some modifications added afterwards to improve its accuracy. The model is based on neural networks. Temperature scenarios obtained from historical data are used as inputs to the neural networks in order to create load scenarios, and these load scenarios are then transformed into quantiles. By using a feature selection approach that is based on a stepwise regression technique, a neural network based model is developed for each zone. Furthermore, a dynamic choice of the temperature scenarios is suggested. The feature selection and dynamic choice of the temperature scenarios can improve the quantile scores considerably, resulting in very accurate forecasts among the top teams.     

Data Envelopment Analysis with Reverse Inputs and Outputs

Data envelopment analysis (DEA) assumes that inputs and outputs are measured on scales in which larger numerical values correspond to greater consumption of inputs and greater production of outputs. We present a class of DEA problems in which one or more of the inputs or outputs are naturally measured on scales in which higher numerical values represent lower input consumption or lower output production. We refer to such quantities as reverse inputs and reverse outputs. We propose to incorporate reverse inputs and outputs into a DEA model by returning to the basic principles that lead to the DEA model formulation. We compare our method to reverse scoring, the most commonly used approach, and demonstrate the relative advantages of our proposed technique. We use this concept to analyze all 30 Major League Baseball (MLB) organizations during the 1999 regular season to determine their on-field and front office relative efficiencies. Our on-field DEA model employs one output and two symmetrically defined inputs, one to measure offense and one to measure defense. The defensive measure is such that larger values correspond to worse defensive performance, rather than better, and hence is a reverse input. The front office model uses one input. Its outputs, one of which is a reverse output, are the inputs to the on-field model. We discuss the organizational implications of our results.     

基于集成学习的多式联运货运量预测模型构建

货运量精准预测是多式联运网络高效协同发展的重要基础,货运量时变性强、数据多样性缺失是实现精准货运量预测的问题所在。基于此,通过挖掘货物运输量(集装箱)的时间变化特征,构建初始相关时间特征输入集,结合斯皮尔曼相关性系数分布,采用Bagging+BP集成学习方法训练多个弱分类器,最终组合获取高精度的强学习模型。以南京龙潭港为例,对自回归移动平均模型(ARIMA)、Bagging+BP集成学习网络以及长短时记忆神经网络(LSTM)三种模型进行评价,实验结果表明,相比于其他模型,提出的Bagging+BP集成学习网络预测性能良好,有一定的实用价值。     

Ensemble RBF modeling technique for quality design

This paper proposes an ensemble radial basis function neural network that selects important RBF subsets based on Pareto chart using Bootstrap samples. Then, the analysis of variance method is used to determine the choice of the unequal/equal weights. The effectiveness of the proposed technique is illustrated with a micro-drilling process. The comparison results show that the proposed technique can not only improve the model prediction performance, but also generate a reliable scheme for quality design.     

Estimating capacity utilization of Chinese manufacturing industries

This study investigates the capacity utilization (CU) of Chinese manufacturing industries, using a CU indicator based on data envelopment analysis and directional distance functions (DDFs). The inputs are separated into variable inputs and a quasi-fixed input to measure the gap of DDFs, which indicated either under-utilization of inputs or overcapacity. Moreover, we define an indicator for CU change over time and introduced the corresponding decomposition. We note that, during the study time period (2007–2010), the CU of Chinese manufacturing industries improved, which implies that Chinese manufacturing industries expanded their production and got closer to their capacity during the examined period. The driving force of this improvement is technical changes. The higher average CU values of light manufacturing industries than that of the heavy industries and the extremely high CU values of two light industries reveal a severe overcapacity problem in the light industries. We also provided the methods and conduct analysis on determining optimal variable inputs and the type of the overcapacity on specific DMUs. The bootstrap regression procedures are employed to test the influence of environmental variables on CU values. Finally, we provide policy implications and suggestions for policymakers who oversee the development of Chinese manufacturing industries.     

中国省际高技术产业效率测度与分解——基于共享投入关联DEA模型

本文基于高技术产业科技成果产出和科技成果转化为生产力两阶段视角,考虑"中间产品产出再投入"和"初始投入在两个子系统间的分配结构",构建共享投入关联DEA模型,利用1999～2010年我国29个省份的面板数据,测算高技术产业系统效率和子系统的纯技术效率,并与关联DEA模型、BCC模型的结果进行比较。结果表明,共享投入关联DEA模型在计算效率水平的同时,还可得到中间产品的转化信息和初始投入的配置信息。     

Modeling emergency evacuation with time and resource constraints: A case study from Gujarat

This study develops an off-site emergency response plan for a nuclear power plant in Gujarat, India subject to time constraints with resource limitations and risk of radiation exposure to victims. We formulate an optimization model to capture the effect of delay in evacuation, limited resource availability, and costs associated with resource allocation. A single chain closed queuing network model with class switching is used to model traffic congestion during evacuation. The throughput measures from the queuing network are used as inputs in the optimization model. Further, two resource allocation strategies are suggested and genetic algorithm is used for optimizing resource utilization and evacuation risk. The results indicate that pooling resources among a cluster of affected areas is most suitable for evacuation. Numerical experiments are conducted to analyze the time trade-offs and the effect of service time variability on the expected evacuation time. The proposed model can serve as an important resource planning and allocation tool for emergency evacuation.     

A coupled Markov chain approach to credit risk modeling

We propose a Markov chain model for credit rating changes. We do not use any distributional assumptions on the asset values of the rated companies but directly model the rating transitions process. The parameters of the model are estimated by a maximum likelihood approach using historical rating transitions and heuristic global optimization techniques.We benchmark the model against a GLMM model in the context of bond portfolio risk management. The proposed model yields stronger dependencies and higher risks than the GLMM model. As a result, the risk optimal portfolios are more conservative than the decisions resulting from the benchmark model.     

Comparing probabilistic forecasts of the daily minimum and maximum temperature

Understanding changes in the frequency, severity, and seasonality of daily temperature extremes is important for public policy decisions regarding heat waves and cold snaps. A heat wave is sometimes defined in terms of both the daily minimum and maximum temperature, which necessitates the generation of forecasts of their joint distribution. In this paper, we develop time series models with the aim of providing insight and producing forecasts of the joint distribution that can challenge the accuracy of forecasts based on ensemble predictions from a numerical weather prediction model. We use ensemble model output statistics to recalibrate the raw ensemble predictions for the marginal distributions, with ensemble copula coupling used to capture the dependency between the marginal distributions. In terms of time series modelling, we consider a bivariate VARMA-MGARCH model. We use daily Spanish data recorded over a 65-year period, and find that, for the 5-year out-of-sample period, the recalibrated ensemble predictions outperform the time series models in terms of forecast accuracy.     

Null models of economic networks: the case of the world trade web

In all empirical-network studies, the observed properties of economic networks are informative only if compared with a well-defined null model that can quantitatively predict the behavior of such properties in constrained graphs. However, predictions of the available null-model methods can be derived analytically only under assumptions (e.g., sparseness of the network) that are unrealistic for most economic networks like the world trade web (WTW). In this paper we study the evolution of the WTW using a recently-proposed family of null network models. The method allows to analytically obtain the expected value of any network statistic across the ensemble of networks that preserve on average some local properties, and are otherwise fully random. We compare expected and observed properties of the WTW in the period 1950–2000, when either the expected number of trade partners or total country trade is kept fixed and equal to observed quantities. We show that, in the binary WTW, node-degree sequences are sufficient to explain higher-order network properties such as disassortativity and clustering-degree correlation, especially in the last part of the sample. Conversely, in the weighted WTW, the observed sequence of total country imports and exports are not sufficient to predict higher-order patterns of the WTW. We discuss some important implications of these findings for international-trade models.     

Parameter-efficient deep probabilistic forecasting

Probabilistic time series forecasting is crucial in many application domains, such as retail, ecommerce, finance, and biology. With the increasing availability of large volumes of data, a number of neural architectures have been proposed for this problem. In particular, Transformer-based methods achieve state-of-the-art performance on real-world benchmarks. However, these methods require a large number of parameters to be learned, which imposes high memory requirements on the computational resources for training such models. To address this problem, we introduce a novel bidirectional temporal convolutional network that requires an order of magnitude fewer parameters than a common Transformer-based approach. Our model combines two temporal convolutional networks: the first network encodes future covariates of the time series, whereas the second network encodes past observations and covariates. We jointly estimate the parameters of an output distribution via these two networks. Experiments on four real-world datasets show that our method performs on par with four state-of-the-art probabilistic forecasting methods, including a Transformer-based approach and WaveNet, on two point metrics (sMAPE and NRMSE) as well as on a set of range metrics (quantile loss percentiles) in the majority of cases. We also demonstrate that our method requires significantly fewer parameters than Transformer-based methods, which means that the model can be trained faster with significantly lower memory requirements, which as a consequence reduces the infrastructure cost for deploying these models.     

Estimating total factor productivity using parametric,nonstochastic cost frontiers

The standard approach to measuring total factor productivity can produce biased results if the data are drawn from a market that is not in long-run competititve equilibrium. This article presents a methodology for adjusting data on output and variable inputs to the values they would have if the market were in long-run competitive equilibrium, given the fixed inputs and input prices. The method uses nonstochastic, parametric translog cost frontiers and calculates equilibrium values for output and varible inputs using an iterative linear programming procedure. Data from seven industries for 1970–1979 are used to illustrate the methodology.The editor for this paper was William H. Greene.     

Correlated daily time series and forecasting in the M4 competition

We participated in the M4 competition for time series forecasting and here describe our methods for forecasting daily time series. We used an ensemble of five statistical forecasting methods and a method that we refer to as the correlator. Our retrospective analysis using the ground truth values published by the M4 organisers after the competition demonstrates that the correlator was responsible for most of our gains over the naïve constant forecasting method. We identify data leakage as one reason for its success, due partly to test data selected from different time intervals, and partly to quality issues with the original time series. We suggest that future forecasting competitions should provide actual dates for the time series so that some of these leakages could be avoided by participants.     

Analysts' utilization of historical earnings information

This study examines whether security analysts (in)efficiently utilize the information contained in past series of annual and quarterly earnings in producing earnings forecasts. To do so, it investigates whether equal-weighted combinations of security analysts' forecasts with forecasts from statistical models based on historical earnings are superior, both in terms of being a better surrogate for the market's expectations of earnings and of accuracy, to forecasts from either one of these two sources. The empirical findings indicate that, although analysts' forecasts are superior to forecasts from statistical models, performance can be improved—both in terms of accuracy and also of being a better surrogate for market earnings expectations—by combining analysts' forecasts with forecasts from statistical models based on past quarterly earnings. Improvements in proxying for market earnings expectations were obtained even when analysts' forecasts made in June of the forecast year were used in the combinations. An implication of these findings is that investors can improve their investment decisions by using an average of the mean analysts' forecasts and the forecast produced by a time-series model of quarterly earnings in their investment decisions.     

Comparing trained and untrained probabilistic ensemble forecasts of COVID-19 cases and deaths in the United States

The U.S. COVID-19 Forecast Hub aggregates forecasts of the short-term burden of COVID-19 in the United States from many contributing teams. We study methods for building an ensemble that combines forecasts from these teams. These experiments have informed the ensemble methods used by the Hub. To be most useful to policymakers, ensemble forecasts must have stable performance in the presence of two key characteristics of the component forecasts: (1) occasional misalignment with the reported data, and (2) instability in the relative performance of component forecasters over time. Our results indicate that in the presence of these challenges, an untrained and robust approach to ensembling using an equally weighted median of all component forecasts is a good choice to support public health decision-makers. In settings where some contributing forecasters have a stable record of good performance, trained ensembles that give those forecasters higher weight can also be helpful.     

Is human and financial investment in Chinese universities effective?

To obtain high-quality human resources and to gain advantages in the context of global competition, countries are constantly promoting the development of higher education. Enormous human and financial resources were invested in this “battle”. However, the “inputs” consumed by universities are largely ignored. A major issue, evaluating how well a country's universities make use of input resources in their output path, is of economic and social significance. The focus of this study is to embed the concept of capacity utilization in education and to assess the university's performance. We construct an attainable output-oriented capacity utilization measurement model by taking into account the heterogeneity of the region. According to the usage of the input, we divide the decision making units into four areas and identify the causes of inefficiency. Thus, the methodology proposed in this study provides policymakers an applicable tool on differences in educational inputs across regions. To illustrate the method, we conduct an empirical analysis using data from 54 universities directly under the Ministry of Education in China.     

Leveraged stock portfolios over long holding periods: A continuous-time model

We use a continuous-time model to derive return and wealth distributions for leveraged portfolios over long holding periods. These theoretical distributions closely match empirical distributions obtained from a resampling procedure. The expected annualized return is a concave function of the degree of leverage. With historical parameter values, the function is maximized at 203% stock, borrowing an amount equal to 103% of net wealth. This maximal stock proportion is considerably reduced if the borrowing rate is higher than the historical lending rate.     

Determining paths of innovation: The role of culture on the adoption on organic farming management

The aim of this paper is to understand whether national culture is linked with operative performance of organic agriculture. The ultimate goal is, indeed, to measure the impact of social change on technical innovation and sustainability issues. We built an econometric model where the impact of some of the six dimensions of national culture by Geert Hofstede is tested on inputs and outputs of organic production, on a country-based scale. We collected data about the evolution of organic agriculture through the last two decades, extracting data about 27 countries of the European Union from 2000 to 2017. We tested the dataset into two passages: we first employed a DEA model to assess efficiency of organic farming, and then we correlated results to Hofstede 6-D model of national culture. We observed the existence of a relationship between culture and innovation, explaining the importance of determined values in people's management within the firm. The operative performance of organic agriculture is linked to determined culturally influenced business values. We used efficiency results as variables for measuring innovation approach based on cultural models. It has been enhanced the relevance to the nexus innovation-culture approaching to one of the most innovation-resistant industrial sectors.