Limited Historical Data Research Articles

Prediction of meteorological drought and standardized precipitation index based on the random forest (RF), random tree (RT), and Gaussian process regression (GPR) models.

Agriculture, meteorological, and hydrological drought is a natural hazard which affects ecosystems in the central India of Maharashtra state. Due to limited historical data for drought monitoring and forecasting available in the central India of Maharashtra state, implementing machine learning (ML) algorithms could allow for the prediction of future drought events. In this paper, we have focused on the prediction accuracy of meteorological drought in the semi-arid region based on the standardized precipitation index (SPI) using the random forest (RF), random tree (RT), and Gaussian process regression (GPR-PUK kernel) models. A different combination of machine learning models and variables has been performed for the forecasting of metrological drought based on the SPI-6 and 12months. Models were developed using monthly rainfall data for the period of 2000-2019 at two meteorological stations, namely, Karanjali and Gangawdi, each representing a geographical region of Upper Godavari river basin area in the central India of Maharashtra state which frequently experiences droughts. Historical data from the SPI from 2000 to 2013 was processed to train the model into machine learning model, and the rest of the 2014 to 2019-year data were used for testing to forecast the SPI and metrological drought. The mean square error (MSE), root mean square error (RMSE), adjustedR2, Mallows' (Cp), Akaike's (AIC), Schwarz's (SBC), and Amemiya's PC were used to identify the best combination input model and best subregression analysis for both stations of SPI-6 and 12. The correlation coefficient ([Formula: see text]), mean absolute error (MAE), root mean square error (RMSE), relative absolute error (RAE), and root relative squared error (RRSE) were used to perform evaluation for SPI-6 and 12months of both stations with RF, RT, and GPR-PUK kernel models during the training and testing scenarios. The results during testing phase revealed that the RF was found as the best model in forecasting droughts with values of [Formula: see text], MAE, RMSE, RAE (%), and RRSE (%) being 0.856, 0.551, 0.718, 74.778, and 54.019, respectively, for SPI-6 while 0.961, 0.361, 0.538, 34.926, and 28.262, respectively, for SPI-12 scales at Gangawdi station. Further, the respective values of evaluators at Karanjali station were 0.913 and 0.966, 0.541 and 0.386, 0.604 and 0.589, 52.592 and 36.959, and 42.315 and 31.394 for PUK kernel and RT models, respectively, during SPI-6 and SPI-12. Machine learning models are potential drought warning techniques because they take less time, have fewer inputs, and are less sophisticated than dynamic or scientific models.

Sales Volume Forecast of Typical Auto Parts Based on BiGRU: A Case Study

Inventory management is an important part of the auto parts supplier business. Accurate prediction of sales volume for different auto parts is the basis for staff to formulate marketing strategies and procurement plans. Based on the limited historical sales data of the South China, North China and East China branches of an auto parts company, some prediction models are trained and tested to determine the best model for predicting future production sales. An orthogonal experimental method is used to implement hyperparameter estimation for the prediction models on this basis. In addition, a posteriori test is used to verify the validity and accuracy of the Bi-GRU model in predicting the sales volume of typical auto parts. The results show that, compared with other models, the bidirectional gated recurrent unit (Bi-GRU) model has the highest accuracy in testing and is used to predict the future sales of typical auto parts. The posterior test proved that the validity and accuracy of the Bi-GRU model is verified. The orthogonal experiment method can effectively realize the hyperparameter estimation for each model. According to the prediction results, the sales volume of blind drive caps in South China, North China and East China will reach 18235, 17030 and 14949 pieces, respectively, after 90 days. Meanwhile, the corresponding sales volume of bolts will reach 13141, 15062 and 10253 pieces, respectively.

Reliable design of a congested disaster relief network: A two-stage stochastic-robust optimization approach

The preparedness of humanitarian relief networks can be enhanced by pre-positioning resources in strategic locations and using them when disasters strike, a strategy that gives rise to a two-stage planning problem. This paper presents a novel two-stage stochastic-robust optimization approach for integrated planning of pre- and post-disaster positioning and allocation of relief resources, while taking into consideration the uncertainty about demand for relief services and disruptions in the relief facilities and the transportation network. The proposed approach enables planners to effectively use limited historical data and imperfect experts’ opinions to obtain robust solutions while avoiding the over-conservatism of classical robust optimization methods. The objective sought is to minimize the expected total time victims need to receive assistance, including both access time to facilities and waiting/service time in them. Congestion in relief facilities is accounted for by modeling them as queuing systems and penalizing waiting time. A decomposition method based on column-and-constraint generation is implemented to solve the problem, whereas the nonlinear terms corresponding to queuing in the second-stage problem are handled using a direct search procedure. Applicability of the proposed approach is demonstrated through a real case study and the numerical results are analyzed to draw managerial insights.

Regional tree-ring oxygen isotope deduced summer monsoon drought variability for Kumaun-Gharwal Himalaya

The precipitation in the Kumaun-Gharwal (Uttarakhand) Himalaya (KGH) is predominantly regulated by the Indian summer monsoon (ISM) and has been declining over the last century. However, because of limited historical data, it is difficult to place this recent decreasing trend in the context of pre-anthropogenic influences and understand a fuller range of hydroclimate scenarios for the region. Thus, we developed a 508-year regional tree-ring stable oxygen isotope (δ18OTR) record for multiple coniferous taxa (Abies spectabilis, Cedrus deodara and Picea smithiana) for the KGH, which spanned 1508–2015 CE. The δ18OTR record explained 35.8% of instrumental June–August Palmer Drought Severity Index (JJAPDSI) variance modelled from a nearby meteorological station. The JJAPDSI reconstruction shows regional coherency with both instrumental and proxy-based independent hydroclimatic records providing confidence that it is a reliable proxy to assess the long term hydroclimatic variability of the region. We identified decreasing strength of JJAPDSI reconstruction in KGH during recent decades indicating intensified drought. This is in the context of pre-anthropogenic influences and understand a fuller range of hydroclimate scenarios for the region. The δ18OTR also has coherency with Pacific and Indian Ocean sea surface temperature variability illustrating the broader teleconnections that influence the regional hydroclimate.

Distributionally Robust Unit Commitment with N-k Security Criterion and Operational Flexibility of CSP

In order to reduce the conservatism of the robust optimization method and the complexity of the stochastic optimization method and to enhance the ability of power systems to deal with occasional line fault disturbance, this paper proposes a distributionally robust unit commitment (DRUC) model with concentrating solar power (CSP) operational flexibility and N-k safety criterion under distributed uncertainty. According to the limited historical sample data, under the condition of satisfying a certain confidence level, based on the imprecise Dirichlet model (IDM), an ambiguity set is constructed to describe the uncertainty of transmission line fault probability. Through the identification of the worst probability distribution in the ambiguity set, the adaptive robust optimal scheduling problem is transformed into a two-stage robust optimization decision model under the condition of deterministic probability distribution. The CSP flexibility column and constraint generation (C&CG) algorithm is used to process the model and the main problem and subproblem are solved by using the Big-M method, linearization technique, and duality principle. Then, a mixed integer linear programming problem (MILP) model is obtained, which effectively reduces the difficulty of solving the model. Finally, case studies on the IEEE 14 bus system and the IEEE 118 bus system demonstrate the efficiency of the proposed method, such as enhancing the ability of power systems to cope with occasional line fault disturbances and reducing the conservatism of the robust optimization method.

Reducing smearing effect in contribution plots and improving fault detection via polynomial approximated isomap embeddings

AbstractDimension reduction is an essential method used in multivariate statistical process monitoring for fault detection and diagnosis. Principal component analysis (PCA) and independent component analysis (ICA) are the most frequently used linear dimensional reduction tools, and the contribution plot is the most popular fault isolation method in the absence of any prior information on the faults. These methods, however, come with their shortcomings. The fault detection capability of linear methods may not be sufficient for non‐linear processes, and smearing effect is known to deteriorate the diagnostics obtained from contribution plots. While the fault detection rate may be increased by kernelized methods or deep artificial neural network models, tuning data‐dependent hyperparameter(s) and network structure with limited historical data is not an easy task. Furthermore, the resulting non‐linear models often do not directly possess fault isolation capability. In the current study, we aim to devise a novel method named ICApIso‐PCA, which offers non‐linear fault detection and isolation in a rather straightforward manner. The rationale of ICApIso‐PCA mainly involves building a non‐linear scores matrix, composed of principal component scores and high‐order polynomial approximated isomap embeddings, followed by implementation of the ICA‐PCA algorithm on this matrix. Applications on a toy dataset and the Tennessee Eastman plant show that the I2 index from ICApIso‐PCA yields a high fault detection rate and offers accurate contribution plots with diminished smearing effects compared to those from traditional monitoring methods. Easy implementation and the potential for future research are further advantages of the proposed method.

Characterization of Air Quality Monitoring Stations in North Region of Malaysia Using Chemometric Technique in Relation to Human Health Impacts

Air quality status plays a vital indicator towards environment, economy as well as human health impacts of a country. Therefore, characterizing the spatial air quality monitoring stations is critical as it can be grouped into clusters in order to minimize the numbers of stations. It could be more efficient towards human health monitoring by strengthening management and reducing costs. However, the assessment of air quality pattern requires multiple variables to be analysed. It becomes a multivariate problem that requires novel methodologies as agglomerative hierarchical cluster analysis (AHC) is used primarily to determine cluster pattern [1]. It becomes a multidimensional problem that necessitates unique approaches, known as hybrid cluster technique. Therefore, this research aimed to show the hybrid cluster framework in air quality monitoring stations in north region of Peninsular Malaysia in order to give a superior geographic cluster distribution with a distinct validation.
 
 The data set for all air quality monitoring stations in north region (14 stations) was obtained from the Department of Environment, Malaysia (DOE) for the years 2018 to 2019 (two years). Six air quality pollutants (PM10, PM2.5, SO2, NO2, O3 and CO) were involved in this study. Before clustering the data, chemometric techniques such as principal component analysis (PCA) was used to summarize the information content in huge data tables in order to gain a better understanding of the variables in order to reduce dimensionality. The AHC was then created using the PCA factor scores. The factor scores were employed in a discriminant analysis (DA) to verify the clusters [2].
 
 Results from PCA factor scores showed 12 out of 14 stations needed to be further investigated using AHC. High Polluted Region (HPR=three stations), Moderate Polluted Region (MPR= four stations) and Low Polluted Region (LPR=five stations) were established from AHC and have the same characteristics as shown in Figure 1. Each class was distinguished using discriminant analysis (DA) for verification. The researchers discovered that each class has its own set of variables. Based on the DA, the confusion matrix for the clusters was 87.66 % indicating a high percentage of correct classification obtained from the validated data set (Table 1).
 The framework presented here is a novel tool for identifying and categorising stations based on air quality pollutants. The hybrid cluster analysis technique used in this study can produce more precise pollutant distributions, which are useful in air pollution research. In addition, the study could help to improve the existing methodology for air quality assessment in a more simplified way and a better evaluation of the air quality status, and thus could become an alternative way for analysis of changes in air quality, especially in the absence or limited historical data, in response to a better and more sustainable indicator in air quality assessment and management.

TLT: Recurrent fine-tuning transfer learning for water quality long-term prediction

The water quality long-term prediction is essential to water environment management decisions. In recent years, although water quality prediction methods based on deep learning have achieved excellent performance in short-term prediction, these methods are unsuitable for long-term prediction because the accumulation use of short-term prediction will easily introduce noise. Furthermore, The long-term prediction task requires a large amount of data to train the model to obtain accurate prediction results. For some monitoring stations with limited historical data, it is challenging to fully exploit the performance of deep learning models. To this end, we introduce a transfer learning framework into water quality prediction to improve the prediction performance in data-constrained scenarios. We propose a deep Transfer Learning based on Transformer (TLT) model to enable time dependency perception and facilitate long-term water quality prediction. In TLT, we innovatively introduce a recurrent fine-tuning transfer learning method, which can transfer the knowledge learned from source monitoring stations to the target station, while preventing the deep learning model from overfitting the source data during the pre-training phase. So, TLT can fully exert the performance of deep learning models with limited samples. We conduct experiments on data from 120 monitoring stations in major rivers and lakes in China to verify the effectiveness of TLT. The results show that TLT can effectively improve the long-term prediction accuracy of four water quality indicators (pH, DO, NH3-N, and CODMn) from monitoring stations with limited samples.

CPV of the Future: AI-Powered Continued Process Verification for Bioreactor Processes.

According to the standard guidelines by the FDA, process validation in biopharma manufacturing encompasses a life cycle consisting of three stages: process design (PD), process qualification (PQ), and continued process verification (CPV). The validity and efficiency of the analytics methods employed during the CPV require extensive knowledge of the process. However, for new processes and new drugs, such knowledge is often not available from Process performance qualification and Validation (PPQV). In this work, the suitability of methods based on machine learning/artificial intelligence (ML/AI) for the CPV applied in bioprocess monitoring and cell physiological control of the yeast Pichia pastoris (Komagataella phaffii) was studied with limited historical data. In particular, the production of recombinant Candida rugosa lipase 1 (Crl1) under hypoxic conditions in fed-batch cultures was considered as a case study. Supervised and unsupervised machine learning models using data from fed-batch bioprocesses with different gene dosage clones under normoxic and hypoxic conditions were evaluated. Firstly, a multivariate anomaly detection (isolation forest) model was applied to the batch phase of the bioprocess. Secondly, a supervised random forest model for prediction of required operator's control actions during the semiautomated fed-batch phase under hypoxic conditions was assessed to maintain the respiratory quotient (RQ) within the desired range for maximizing the specific production rate (qP ). The performance of these models was tested on historical data using independent evaluation of the process by the process control engineer (subject matter expert-SME), and on real-time data in the case of manual action prediction, where the model was implemented to guide the control of the bioprocess. The work presented here constitutes a proof-of-concept that multivariate analytics methods, based on machine learning, can be a valuable tool for real-time monitoring and control of biopharma manufacturing bioprocesses to improve its efficiency and to assure product quality.

Estimating tropical cyclone-induced wind, waves, and surge: A general methodology based on representative tracks

Tropical Cyclones (TCs) are singular storms causing intense wind, large waves, extreme water levels, and heavy rainfall. TCs prove every year to be one of the most destructive natural phenomena worldwide. The quantitative assessment of the hazards resulting from TCs (i.e., flooding and extreme winds) is challenging since satellite data are only available for recent decades, whereas older historical observations are incomplete and less accurate. In addition, long-term prediction through numerical weather forecasting is still limited. This often results in large uncertainties in the definition of TC hazards associated with events with longer return periods or in areas infrequently impacted by TCs. Even when this information is available, for example through statistical sampling of synthetic TC tracks, the numerical modelling of the associated hazards for all the different TC conditions can lead to computational costs which are often infeasible. Several methodologies that overcome the issues of accuracy and computational efficiency currently exist, but these are not generically applicable, and they tend to focus on specific areas only, for example where TCs typically make landfall. The main contribution of this paper is a novel methodology for the estimation and analysis of TC hydro-meteorological conditions and induced hazards. The method is generically applicable and maximizes accuracy while accounting for computational efficiency. Our approach identifies a smaller but representative set of TC tracks (RTCs) that preserves the information about extremes and the frequency of events of the larger population. The method is successfully applied and validated in a case study in the Bay of Bengal, using a set of synthetic TC tracks representing 1000 years of TC climate. For the best-performing configuration, the required number of scenarios and associated computational costs were reduced by 90% while maintaining accuracy in the simulated offshore storm surges, significant wave height, and windspeeds typically within 10% of the prediction based on the original full set of scenarios. This method is globally applicable and greatly improves the efficiency of TC-related hazard estimation, making it particularly valuable for areas with limited historical data.

Hybrid Deep Learning Applied on Saudi Smart Grids for Short-Term Load Forecasting

Despite advancements in smart grid (SG) technology, effective load forecasting utilizing big data or large-scale datasets remains a complex task for energy management, planning, and control. The Saudi SGs, in alignment with the Saudi Vision 2030, have been envisioned as future electrical grids with a bidirectional flow of power and data. To that end, data analysis and predictive models can enhance Saudi SG planning and control via artificial intelligence (AI). Recently, many AI methods including deep learning (DL) algorithms for SG applications have been published in the literature and have shown superior time series predictions compared with conventional prediction models. Current load-prediction research for the Saudi grid focuses on identifying anticipated loads and consumptions, on utilizing limited historical data and the behavior of the load’s consumption, and on conducting shallow forecasting models. However, little scientific proof on complex DL models or real-life application has been conducted by researchers; few articles have studied sophisticated large-scale prediction models for Saudi grids. This paper proposes hybrid DL methods to enhance the outcomes in Saudi SG load forecasting, to improve problem-relevant features, and to accurately predict complicated power consumption, with the goal of developing reliable forecasting models and of obtaining knowledge of the relationships between the various features and attributes in the Saudi SGs. The model in this paper utilizes a real dataset from the Jeddah and Medinah grids in Saudi Arabia for a full year, 2021, with a one-hour time resolution. A benchmark strategy using different conventional DL methods including artificial neural network, recurrent neural network (RNN), conventional neural networks (CNN), long short-term memory (LSTM), gated recurrent unit (GRU), and different real datasets is used to verify the proposed models. The prediction results demonstrate the effectiveness of the proposed hybrid DL models, with CNN–GRU and CNN–RNN with NRMSE obtaining 1.4673% and 1.222% improvements, respectively, in load forecasting accuracy.

Transfer Learning Featured Short-Term Combining Forecasting Model for Residential Loads With Small Sample Sets

In recent years, a large variety of short-term load forecasting methodologies have been proposed, but a common drawback of them is that require a large amount of historical data to train the model. However, in reality, the cases in limited historical data would be more frequently occurred, for example, a new house or substation is built. To tackle this issue, a combining forecasting method of deep residual neural network (ResNet)-based transfer learning for residential loads is proposed in this article. To improve the accuracy and reliability of transfer learning, a novel deep ResNet with dual skip connections (ResNet-DSC) is proposed as the base model. Then with sparse training data, a Bayesian probability weighted averaging (approach is proposed, to address the model combination parameters estimation problem. In addition, probability density function forecasting is also delivered by the method. For demonstration, a series of experiments, with actual residential load data, was performed. With the comparison with other transfer learning and nontransfer learning approaches, the effectiveness and improvement of the proposed method have been validated.

Improved Modeling of Kinematics-Induced Geometric Variations in Extrusion-Based Additive Manufacturing Through Between-Printer Transfer Learning

For extrusion-based additive manufacturing, the variation in material deposition can significantly affect printed material distribution, causing infill nonuniformity and defects. These variations are induced by kinematic variations of the printer extruder. Such infill nonuniformity is more significant in an application of collaborative printing systems by which multiple printers’ extrudes co-create the same structure since more accelerate–decelerate kinematic cycles are involved. There is a lack of a quantitative understanding of the impact of printing kinematics on such variations to guide the printing process control. This article deals with the challenge by establishing a mathematical model that quantifies the printing width variations along the printing paths induced by printing speed and acceleration. The model provides vital information for predicting infill pattern nonuniformity and potentially enables using G-code adjustment to compensate for the infill errors in future research. In addition, since the model captures the mechanism of kinematics-induced variations, it provides a way of between-printer knowledge transfer on estimating printing errors. This article further proposes an informative-prior-based transfer learning algorithm to improve the quality prediction model for a printer with limited historical data by leveraging the shared data from interconnected 3-D printers. A case study based on experiments validated the effectiveness of the proposed methodology. <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Note to Practitioners</i> —This article quantitatively studies the impact of extruder kinematics on geometric variations and printing quality in extrusion-based 3-D printing processes. The model can help predict the geometric printing quality and related defects, such as overfill or underfill problems given kinematics setup by G-code. This study can expedite the learning process of printing variations induced by kinematics for new printers to set up monitoring and G-code adjustment for process control in the early stage of production when the data are limited. In the long run, such between-printer transfer learning has the potential to enable the transfer learning for interconnected collaborative 3-D printing systems with improved printing efficiency and quality.

Data-driven Wasserstein distributionally robust mitigation and recovery against random supply chain disruption

This paper studies joint robust network design and recovery investment management in a production supply chain, considering limited historical data about disruptions and their possibilities. The supply chain is subject to uncertain disruptions that reduce production capacity at plants, and the cascading failures propagate along the supply chain network. A data-driven two-stage distributionally robust optimization model with Wasserstein ambiguity set (TWDRO) is constructed to determine the strategic location and tactical allocation decisions in the first stage as well as the operational production and inventory decisions and recovery policy by recourse in the second stage. This paper also proposes a model for depicting a recovery-fund based mitigation strategy, and the model is general in depicting the accelerated, constant-speed, and decelerated recovery processes. In addition, partial backorder policy is adopted to depict the customers’ choices upon product stockout. The TWDRO model is solved by converting it to a mixed integer linear programming model and designing a joint solution method using benders decomposition and genetic algorithm. The performance of TWDRO solutions is demonstrated through numerical experiments and a case study, benchmarking on the stochastic programming and robust optimization approaches. This paper shows the effectiveness and robustness of TWDRO and provides managerial implications and suggestions for supply chain disruption and recovery management.

Improved scalability and risk factor proxying with a two-step principal component analysis for multi-curve modelling

We consider the practice-relevant problem of modelling multiple price curves to support activities such as price curve simulation and risk management. In this multi-curve setting, the challenge is to jointly capture the risk-factor relationships within each curve and the risk-factor relationships between the curves. Contributing to the existing literature, we develop a novel two-step Principal Component Analysis (PCA) method, which we label PCA2, that addresses this challenge. The concept of PCA2 first derives components describing the dynamics of each curve, and then, second, combines these to describe the dynamics across all the curves. The benefits of PCA2 over PCA are: (i) improved scalability allowing for greater computational efficiency and smaller data structures rendering multi-threading more feasible; (ii) components that remain identifiable at the curve level; and (iii) leveraging the last property, PCA2, unlike PCA, offers the capability of proxying new curves for which limited historical data exists, using the first-step components from a related curve and estimating second-level correlations empirically. PCA2 is a novel multi-curve modelling approach that will appeal, for these reasons, to many practitioners, especially those working in risk management.

A comparative analysis of data mining techniques for agricultural and hydrological drought prediction in the eastern Mediterranean

Drought is a natural hazard which affects ecosystems in the eastern Mediterranean. However, limited historical data for drought monitoring and forecasting are available in the eastern Mediterranean. Thus, implementing machine learning (ML) algorithms could allow for the prediction of future drought events. In this context, the main goals of this research were to capture agricultural and hydrological drought trends by using the Standardized Precipitation Index (SPI) and to assess the applicability of four ML algorithms (bagging (BG), random subspace (RSS), random tree (RT), and random forest (RF)) in predicting drought events in the eastern Mediterranean based on SPI-3 and SPI-12. The results reveal that hydrological drought (SPI-12, −24) was more severe over the study area, where most stations showed a significant (p < 0.05) negative trend. The accuracy of ML algorithms in drought prediction varied in relation to the implementation stage. In the training stage, RT outperformed the other algorithms (Root mean square error (RMSE) = 0.3, Correlation Coefficient (r) = 0.97); the performance of the algorithms can be ranked as follows: RT > RF > BG > RSS for both SPI-3 and SPI-12. In the testing stage, both the BG and RF algorithms had the highest correlation r(observed vs. predicted) (0.58–0.64) and lowest RMSE (0.68–0.88). In contrast, the lowest correlation r(observed vs. predicted) (0.3–0.41) and highest RMSE (0.94–1.10) was calculated for the RT algorithm. However, BG was more dynamic in drought capturing, with the lowest RMSE and highest correlation. In the validation stage, the BG performance was satisfactory (RMSE = 0.62–0.83, r = 0.58–0.79). The output of this research will help decision-makers with drought mitigation plans by using the new four machine learning algorithms.

Uncovering forest dynamics using historical forest inventory data and Landsat time series

While many countries regularly produce timely and accurate estimates for a range of forest attributes through national forest inventory programs, many have access to no such inventory or their inventory is discontinuous in space and time, such as a sample-based approach in Ukraine limited to two regional (Ivano-Frankivsk and Sumy) inventories conducted in 2008–2015 at a total area of 37,700 ha. This study addresses the extent to which limited historical forest inventory data may facilitate forest mapping efforts using dense Landsat time series (LTS) and mapping techniques (i.e., classification and imputation) to predict multiyear forest dynamics. We used the Continuous Change Detection and Classification (CCDC) segmentation approach to extract inter- and intra-annual trends in LTS and utilized these fitted trends as covariates in further modeling. We developed random forest (RF) classification models for forest mapping based on field data collection and LTS, and then generated yearly forest maps for 1990–2020. The RF model accuracies were high for both producer’s and user’s accuracy (>0.93 ± 0.04). Based on yearly forest dynamics, we detected an increasing trend in forest loss during 2000–2010 and 2010–2020 in our regions. We used the gradient nearest neighbor (GNN) method to generate species presence (basal area (BA) ≥ 1.0 m2 ha−1) and growing stock volume (GSV) maps at the 30-m pixel level. Our nearest neighbor imputation models achieved better accuracies (Cohen’s kappa > 0.4) for species that are prevalent within regions, occupy distinct geographical areas, and have higher plot BA abundance. To obtain estimates of species abundance with higher R2, we used the GNN model with k = 3 nearest neighbors. Aggregating species based on similar environmental niche resulted in greater R2 values of BA that varied within two regions from low (0.129 and 0.416) for a group of hardwood deciduous species to high (0.531 and 0.712) for coniferous species. The accuracy was systematically better at 5-km hexagonal level, thus we recommend using the maps at coarser aggregations. The study showed that forests accumulated more GSV in 1990–2000, albeit higher rates of GSV loss were observed during the two last decades. We attribute the spatial and temporal character of forest change to forestry activities in the past, the current age structure of forests, and afforestation of abandoned farmlands. We are aware that forecasting performance of the nearest neighbor imputation approach merits more detailed consideration as new forest inventory data are available.

Short-term trajectory prediction for individual metro passengers integrating diverse mobility patterns with adaptive location-awareness

Short-term trajectory prediction (StTP) for individual metro passengers is of great importance in intelligent transportation systems and real-time security risk management. Existing research efforts still have shortcomings in constructing an interpretable StTP model with limited historical trajectory data for individual metro passengers. This paper thus intends to enhance the existing StTP methods in two aspects: (1) encoding the trajectory sequences with universal and personalized mobility rules to explain the mobility patterns of metro passengers, and (2) constructing an interpretable StTP model with adaptive location-awareness to improve the prediction accuracy. Along this line, this study develops a novel framework for individual metro passengers by integrating diverse mobility patterns with adaptive location-awareness (StTP-ML). Particularly, a trajectory representation approach is proposed to process the observed Wi-Fi probe data and transfer them into structured trajectories. Then, the diverse mobility patterns of individual passengers are discovered, including temporal periodicity, spatial symmetry, and sequence correlation. Finally, an adaptive location-aware model integrated with individual diverse mobility patterns-priority strategy is constructed. To evaluate the proposed StTP-ML model, this research also conducts experiments on real-world passenger trajectory data at a busy metro station in Shanghai, China. The experimental results show that the average accuracy rate of the StTP-ML-E model is improved up to 84%, which outperforms all other baseline models in the test. It demonstrates that the proposed StTP-ML model can integrate more interpretable mobility patterns than the baseline models and provide more accurate trajectories for individual metro passengers in future time periods.

Data-driven robust dual-sourcing inventory management under purchase price and demand uncertainties

We develop an actionable data-driven approach to a periodic-review dual sourcing inventory management system in the presence of purchase price and demand uncertainties. The two supply sources differ in their lead times and prices due to, e.g., different transportation modes. We adopt robust optimization because the limited historical data is insufficient to construct meaningful distributions to characterize purchase price and demand fluctuations. Specifically, we build a robust rolling-horizon model and, in particular, construct the uncertainty sets from data and business insights. Using a real four-year data set, we show that our approach can yield significant cost savings compared to the other popular methods. Our experiments echo the earlier theoretical finding that a firm may incur a lower cost under a more volatile purchase price process. However, we find that under data-driven decision-making, in comparison with the theoretical results that assume complete distributional information, some interesting results arise. For example, first, considering a longer planning horizon may backfire. Second, some feasible region-reducing business constraints such as limited inventory capacity may lead to unintended benefits. These are consequences of protecting model performance from sampling error and our practically limited forecast ability, as almost always, to characterize uncertainties. Our research therefore calls for prudence in extending theoretical insights to data-driven decision-making scenarios.

Reality Model-Based Facility Management Framework for Existing Building

Facility management (FM) is the most extended phase of the entire life cycle of buildings and incurs the highest expenses. However, limited quality drawings and historical data make information flow more complicated for many existing buildings and lead to inefficient FM. This research proposes a systematic and comprehensive reality point cloud model-based framework to support the FM process with feasible visual data collection. Specifically, this framework covers an end-to-end workflow of using a reality model to implement FM strategies, including visual data capture process for generating a reality model; FM information attachment in a reality model for issue queries; and information flow for issue tracking. The resulting reality model captures with critical and necessary FM data input can be used as an asset information model (AIM). We have also defined stakeholders' responsibilities in implementing FM and how the data are transferred, updated, and used within a reality point cloud-based AIM by stakeholders. The framework is evaluated by interviewing FM experts, and the results demonstrate that implementing the framework could improve the FM process's efficiency by using reality models.