Forecasting, Volume 4, Issue 1 (March 2022) – 22 articles

We propose a functional time series method to obtain accurate multi-step-ahead forecasts for age-specific mortality rates. The dynamic functional principal component analysis method is used to decompose the mortality curves into dynamic functional principal components and their associated principal component scores. Machine-learning-based multi-step-ahead forecasting strategies, which automatically learn the underlying structure of the data, are used to obtain the future realization of the principal component scores. The forecasted mortality curves are obtained by combining the dynamic functional principal components and forecasted principal component scores. The point and interval forecast accuracy of the proposed method is evaluated using six age-specific mortality datasets and compared favorably with four existing functional time series methods. Full article

Demographic change is leading to the aging of German society. As long as the baby boom cohorts are still of working age, the working population will also age—and decline as soon as this baby boom generation gradually reaches retirement age. At the same time, there has been a trend toward increasing absenteeism (times of inability to work) in companies since the zero years, with the number of days of absence increasing with age. We present a novel stochastic forecast approach that combines population forecasting with forecasts of labor force participation trends, considering epidemiological aspects. For this, we combine a stochastic Monte Carlo-based cohort-component forecast of the population with projections of labor force participation rates and morbidity rates. This article examines the purely demographic effect on the economic costs associated with such absenteeism due to the inability to work. Under expected future employment patterns and constant morbidity patterns, absenteeism is expected to be close to 5 percent by 2050 relative to 2020, associated with increasing economic costs of almost 3 percent. Our results illustrate how strongly the pronounced baby boom/baby bust phenomenon determines demographic development in Germany in the midterm. Full article

Effective weather forecast information helps smallholder farmers improve their adaptation to climate uncertainties and crop productivity. The main objective of this study was to assess the impact of weather forecast adoption on crop productivity. We coupled agent-based and crop productivity models to study the impact of farmers’ management decisions on maize productivity under different rainfall scenarios in Ethiopia. A household survey was conducted with 100 households from 5 villages and was used to validate the crop model. The agent-based model (ABM) analyzed the farmers’ behaviors in crop management under different dry, wet, and normal rainfall conditions. ABM results and crop data from the survey were then used as input data sources for the crop model. Our results show that farming decisions based on weather forecast information improved yield productivity from 17% to 30% under dry and wet seasons, respectively. The impact of adoption rates due to farmers’ intervillage interactions, connections, radio, agriculture extension services, and forecast accuracy brought additional crop yields into the Kebele compared to non-forecast users. Our findings help local policy makers to understand the impact of the forecast information. Results of this study can be used to develop agricultural programs where rainfed agriculture is common. Full article

This work proposes and evaluates a method for the nowcasting of solar irradiance variability in multiple time horizons, namely 5, 10, and 15 min ahead. The method is based on a Convolutional Neural Network structure that exploits infrared sky images acquired through an All-Sky Imager to estimate the range of possible values that the Clear-Sky Index will possibly assume over a selected forecast horizon. All data available, from the infrared images to the measurements of Global Horizontal Irradiance (necessary in order to compute Clear-Sky Index), are acquired at SolarTech^LAB in Politecnico di Milano. The proposed method demonstrated a discrete performance level, with an accuracy peak for the 5 min time horizon, where about 65% of the available samples are attributed to the correct range of Clear-Sky Index values. Full article

Many people do not possess the necessary savings to deal with unexpected financial events. People’s biases play a significant role in their ability to forecast future financial shocks: they are typically overoptimistic, present-oriented, and generally underestimate future expenses. The purpose of this study is to investigate how varying risk information influences people’s financial awareness, in order to reduce the chance of a financial downfall. Specifically, we contribute to the literature by exploring the concept of ‘nudging’ and its value for behavioural changes in personal financial management. While of great practical importance, the role of nudging in behavioural financial forecasting research is scarce. Additionally, the study steers away from the standard default choice architecture nudge, and adds originality by focusing on eliciting implementation intentions and precommitment strategies as types of nudges. Our experimental scenarios examined how people change their financial projections in response to nudges in the form of new information on relevant risks. Participants were asked to forecast future expenses and future savings. They then received information on potential events identified as high-risk, low-risk or no-risk. We investigated whether they adjusted their predictions in response to various risk scenarios or not and how such potential adjustments were affected by the information given. Our findings suggest that the provision of risk information alters financial forecasting behaviour. Notably, we found an adjustment effect even in the no-risk category, suggesting that governments and institutions concerned with financial behaviour can increase financial awareness merely by increasing salience about possible financial risks. Another practical implication relates to splitting savings into different categories, and by using different wordings: A financial advisory institution can help people in their financial behaviour by focusing on ‘targets’, and by encouraging (nudging) people to make breakdown forecasts rather than general ones. Full article

In this paper, we conduct a thorough investigation of the predictive ability of forward and backward stepwise regressions and hidden Markov models for the futures returns of several commodities. The predictive performance relative a standard AR(1) benchmark is assessed under both statistical and economic loss functions. We find that the evidence that either stepwise regressions or hidden Markov models may outperform the benchmark under standard statistical loss functions is rather weak and limited to low-volatility regimes. However, a mean-variance investor that adopts flexible forecasting models (especially stepwise predictive regressions) when building her portfolio, achieves large benefits in terms of realized Sharpe ratios and mean-variance utility compared to an investor employing AR(1) forecasts. Full article

Image classification is widely used to build predictive models for breast cancer diagnosis. Most existing approaches overwhelmingly rely on deep convolutional networks to build such diagnosis pipelines. These model architectures, although remarkable in performance, are black-box systems that provide minimal insight into the inner logic behind their predictions. This is a major drawback as the explainability of prediction is vital for applications such as cancer diagnosis. In this paper, we address this issue by proposing an explainable machine learning pipeline for breast cancer diagnosis based on ultrasound images. We extract first- and second-order texture features of the ultrasound images and use them to build a probabilistic ensemble of decision tree classifiers. Each decision tree learns to classify the input ultrasound image by learning a set of robust decision thresholds for texture features of the image. The decision path of the model predictions can then be interpreted by decomposing the learned decision trees. Our results show that our proposed framework achieves high predictive performance while being explainable. Full article

Temperature is the most used meteorological variable for a large number of applications in urban resilience planning, but direct measurements using traditional sensors are not affordable at the usually required spatial density. On the other hand, spaceborne remote sensing provides surface temperatures at medium to high spatial resolutions, almost compatible with the needed requirements. However, in this case, limitations are represented by cloud conditions and passing times together with the fact that surface temperature is not directly comparable to air temperature. Various methodologies are possible to take benefits from both measurements and analysis methods, such as direct assimilation in numerical models, multivariate analysis, or statistical interpolation. High-resolution thermal fields in the urban environment are also obtained by numerical modelling. Several codes have been developed to resolve at some level or to parameterize the complex urban boundary layer and are used for research and applications. Downscaling techniques from global or regional models offer another possibility. In the Milan metropolitan area, given the availability of both a high-quality urban meteorological network and spaceborne land surface temperatures, and also modelling and downscaling products, these methods can be directly compared. In this paper, the comparison is performed using: the ClimaMi Project high-quality data set with the accurately selected measurements in the Milan urban canopy layer, interpolated by a cokriging technique with remote-sensed land surface temperatures to enhance spatial resolution; the UrbClim downscaled data from the reanalysis data set ERA5; a set of near-surface temperatures produced by some WRF outputs with the building environment parameterization urban scheme. The comparison with UrbClim and WRF of the cokriging interpolated data set, mainly based on the urban canopy layer measurements and covering several years, is presented and discussed in this article. This comparison emphasizes the primary relevance of surface urban measurements and highlights discrepancies with the urban modelling data sets. Full article

As the availability of big data-sets becomes more widespread so the importance of motif (or repeated pattern) identification and analysis increases. To date, the majority of motif identification algorithms that permit flexibility of sub-sequence length do so over a given range, with the restriction that both sides of an identified sub-sequence pair are of equal length. In this article, motivated by a better localised representation of variations in time series, a novel approach to the identification of motifs is discussed, which allows for some flexibility in side-length. The advantages of this flexibility include improved recognition of localised similar behaviour (manifested as motif shape) over varying timescales. As well as facilitating improved interpretation of localised volatility patterns and a visual comparison of relative volatility levels of series at a globalised level. The process described extends and modifies established techniques, namely SAX, MDL and the Matrix Profile, allowing advantageous properties of leading algorithms for data analysis and dimensionality reduction to be incorporated and future-proofed. Although this technique is potentially applicable to any time series analysis, the focus here is financial and energy sector applications where real-world examples examining S&P500 and Open Power System Data are also provided for illustration. Full article

We introduce a simple extension to the CBDX model to project cohort mortality rates to extreme old age. The proposed approach fits a polynomial to a sample of age effects, uses the fitted polynomial to project the age effects to ages beyond the sample age range, then splices the sample and projected age effects, and uses the spliced age effects to obtain mortality rates for the higher ages. The proposed approach can be used to value financial instruments such as life annuities that depend on projections of extreme old age mortality rates. Full article

Supply Chain Finance (SCF) has gradually taken on digital characteristics with the rapid development of electronic information technology. Business audit information has become more abundant and complex, which has increased the efficiency and increased the potential risk of commercial banks, with credit risk being the biggest risk they face. Therefore, credit risk assessment based on the application of digital SCF is of great importance to commercial banks’ financial decisions. This paper uses a hybrid Extreme Gradient Boosting Multi-Layer Perceptron (XGBoost-MLP) model to assess the credit risk of Digital SCF (DSCF). In this paper, 1357 observations from 85 Chinese-listed SMEs over the period 2016–2019 are selected as the empirical sample, and the important features of credit risk assessment in DSCF are automatically selected through the feature selection of the XGBoost model in the first stage, then followed by credit risk assessment through the MLP in the second stage. Based on the empirical results, we find that the XGBoost-MLP model has good performance in credit risk assessment, where XGBoost feature selection is important for the credit risk assessment model. From the perspective of DSCF, the results show that the inclusion of digital features improves the accuracy of credit risk assessment in SCF. Full article

One of the most critical issues for wind energy exploitation is the high variability of the resource, resulting in very difficult forecasting of the power that wind farms can grant. A vast literature has therefore been devoted to wind speed and wind power quantitative forecasting, using different techniques. The widely used statistical and learning models that are based on a continuation in the future of the series’ past behaviour offer a performance that may be much less satisfactory when the values suddenly change their trend. The application to wind speed data of two techniques usually employed for the technical analysis of financial series–namely support and resistances identification and candlestick charts–is investigated here, with the main aim to detect inversion points in the series. They are applied to wind speed data series for two locations in Spain and Italy. The proposed indicators confirm their usefulness in identifying peculiar behaviours in the system and conditions where the trend may be expected to change. This additional information offered to the forecasting algorithms may also be included in innovative approaches, e.g., based on machine learning. Full article

Electrical load forecasting study is required in electric power systems for different applications with respect to the specific time horizon, such as optimal operations, grid stability, Demand Side Management (DSM) and long-term strategic planning. In this context, machine learning and data analytics models represent a valuable tool to cope with the intrinsic complexity and especially design future demand-side advanced services. The main novelty in this paper is that the combination of a Recurrent Neural Network (RNN) and Principal Component Analysis (PCA) techniques is proposed to improve the forecasting capability of the hourly load on an electric power substation. A historical dataset of measured loads related to a 33/11 kV MV substation is considered in India as a case study, in order to properly validate the designed method. Based on the presented numerical results, the proposed approach proved itself to accurately predict loads with a reduced dimensionality of input data, thus minimizing the overall computational effort. Full article

Timely generation of accurate and reliable forecasts of flash flood events is of paramount importance for flood early warning systems in urban areas. Although physically based models are able to provide realistic reproductions of fast-developing inundation maps in high resolutions, the high computational demand of such hydraulic models makes them difficult to be implemented as part of real-time forecasting systems. This paper evaluates the use of a hybrid machine learning approach as a surrogate of a quasi-2D urban flood inundation model developed in PCSWMM for an urban catchment located in Toronto (Ontario, Canada). The capability to replicate the behavior of the hydraulic model was evaluated through multiple performance metrics considering error, bias, correlation, and contingency table analysis. Results indicate that the surrogate system can provide useful forecasts for decision makers by rapidly generating future flood inundation maps comparable to the simulations of physically based models. The experimental tool developed can issue reliable alerts of upcoming inundation depths on traffic locations within one to two hours of lead time, which is sufficient for the adoption of important preventive actions. These promising outcomes were achieved in a deterministic setup and use only past records of precipitation and discharge as input during runtime. Full article

The impact of climate change on soil temperatures at Kuujjuaq, Quebec in northern Canada is assessed. First, long-term historical soil temperature records (1967–1995) are statistically analyzed to provide a climatological baseline for soils at 5 to 150 cm depths. Next, the nature of the relationship between atmospheric variables and soil temperature are determined using a statistical downscaling model (SDSM) and National Centers for Environmental Prediction (NCEP), a climatological data set. SDSM was found to replicate historic soil temperatures well and used to project soil temperatures for the remainder of the century using climate model output Canadian Second Generation Earth System Model (CanESM2). Three Representative Concentration Pathway scenarios (RCP 2.6, 4.5 and 8.5) were used from the Intergovernmental Panel on Climate Change (IPCC) Fifth Assessment Report (AR5). This study found that the soil temperature at this location may warm at 0.9 to 1.2 °C per decade at various depths. Annual soil temperatures at all depths are projected to rise to above 0 °C for the 1997–2026 period for all climate scenarios. The melting soil poses a hazard to the airport infrastructure and will require adaptation measures. Full article

Accurate forecasts of the number of newly infected people during an epidemic are critical for making effective timely decisions. This paper addresses this challenge using the SIMLR model, which incorporates machine learning (ML) into the epidemiological SIR model. For each region, SIMLR tracks the changes in the policies implemented at the government level, which it uses to estimate the time-varying parameters of an SIR model for forecasting the number of new infections one to four weeks in advance. It also forecasts the probability of changes in those government policies at each of these future times, which is essential for the longer-range forecasts. We applied SIMLR to data from in Canada and the United States, and show that its mean average percentage error is as good as state-of-the-art forecasting models, with the added advantage of being an interpretable model. We expect that this approach will be useful not only for forecasting COVID-19 infections, but also in predicting the evolution of other infectious diseases. Full article

The present paper considers the problem of choosing among a collection of competing electricity price forecasting models to address a stochastic decision-making problem. We propose an event-based evaluation framework applicable to any optimization problem, where uncertainty is captured through ensembles. The task of forecast evaluation is simplified from assessing a multivariate distribution over prices to assessing a univariate distribution over a binary outcome directly linked to the underlying decision-making problem. The applicability of our framework is demonstrated for two exemplary profit-maximization problems of a risk-neutral energy trader, (i) the optimal operation of a pumped-hydro storage plant and (ii) the optimal trading of subsidized renewable energy in Germany. We compare and contrast the approach with the full probabilistic and profit–loss-based evaluation frameworks. It is concluded that the event-based evaluation framework more reliably identifies economically equivalent forecasting models, and in addition, the results suggest that an event-based evaluation specifically tailored to the rare event is crucial for decision-making problems linked to rare events. Full article

Vietnam has experienced a tourism expansion over the last decade, proving itself as one of the top tourist destinations in Southeast Asia. The country received more than 18 million international tourists in 2019, compared to only 1.5 million twenty-five years ago. Tourist spending has translated into rising employment and incomes for Vietnam’s tourism sector, making it the key driver to the socio-economic development of the country. Following the COVID-19 pandemic, only 3.8 million international tourists visited Vietnam in 2020, plummeting by 78.7% year-on-year. The latest outbreak in early summer 2021 made the sector continue to hit bottom. Although Vietnam’s tourism has suffered extreme losses, once the contagion is under control worldwide, the number of international tourists to Vietnam is expected to rise again to reach pre-pandemic levels in the next few years. First, the paper aims to provide a summary of Vietnam’s tourism characteristics with a special focus on international tourists. Next, the predictive capability of artificial neural network (ANN) methodology is examined with the datasets of international tourists to Vietnam from 2008 to 2020. Some ANN architectures are experimented with to predict the monthly number of international tourists to the country, including some lockdown periods due to the COVID-19 pandemic. The results show that, with the correct selection of ANN architectures and data from the previous 12 months, the best ANN models can be forecast for next month with a MAPE between 7.9% and 9.2%. As the method proves its forecasting accuracy, it would serve as a valuable tool for Vietnam’s policymakers and firm managers to make better investment and strategic decisions. Full article

The evolution of functional autonomy loss leads to institutionalization of people affected by Alzheimer’s disease (AD), to an alteration of their quality of life and that of their caregivers. To predict loss of functional autonomy could optimize prevention strategies, aids and cost of care. The aim of this study was to develop and to cross-validate a model to predict loss of functional autonomy as assessed by Instrumental Activities of Daily Living (IADL) score. Outpatients with probable AD and with 2 or more visits to the Clinical and Research Memory Centre of the University Hospital were included. Four Tree-Augmented Naïve bayesian networks (6, 12, 18 and 24 months of follow-up) were built. Variables included in the model were demographic data, IADL score, MMSE score, comorbidities, drug prescription (psychotropics and AD-specific drugs). A 10-fold cross-validation was conducted to evaluate robustness of models. The study initially included 485 patients in the prospective cohort. The best performance after 10-fold cross-validation was obtained with the model able to predict loss of functional autonomy at 18 months (area under the curve of the receiving operator characteristic curve = 0.741, 27% of patients misclassified, positive predictive value = 77% and negative predictive value = 73%). The 13 variables used explain 41.6% of the evolution of functional autonomy at 18 months. A high-performing predictive model of AD evolution of functional autonomy was obtained. An external validation is needed to use the model in clinical routine so as to optimize the patient care. Full article

Hybrid methods have been shown to outperform pure statistical and pure deep learning methods at forecasting tasks and quantifying the associated uncertainty with those forecasts (prediction intervals). One example is Exponential Smoothing Recurrent Neural Network (ES-RNN), a hybrid between a statistical forecasting model and a recurrent neural network variant. ES-RNN achieves a 9.4% improvement in absolute error in the Makridakis-4 Forecasting Competition. This improvement and similar outperformance from other hybrid models have primarily been demonstrated only on univariate datasets. Difficulties with applying hybrid forecast methods to multivariate data include (i) the high computational cost involved in hyperparameter tuning for models that are not parsimonious, (ii) challenges associated with auto-correlation inherent in the data, as well as (iii) complex dependency (cross-correlation) between the covariates that may be hard to capture. This paper presents Multivariate Exponential Smoothing Long Short Term Memory (MES-LSTM), a generalized multivariate extension to ES-RNN, that overcomes these challenges. MES-LSTM utilizes a vectorized implementation. We test MES-LSTM on several aggregated coronavirus disease of 2019 (COVID-19) morbidity datasets and find our hybrid approach shows consistent, significant improvement over pure statistical and deep learning methods at forecast accuracy and prediction interval construction. Full article