Hierarchical Prediction Research Articles

A Hierarchical RF-XGBoost Model for Short-Cycle Agricultural Product Sales Forecasting.

Short-cycle agricultural product sales forecasting significantly reduces food waste by accurately predicting demand, ensuring producers match supply with consumer needs. However, the forecasting is often subject to uncertain factors, resulting in highly volatile and discontinuous data. To address this, a hierarchical prediction model that combines RF-XGBoost is proposed in this work. It adopts the Random Forest (RF) in the first layer to extract residuals and achieve initial prediction results based on correlation features from Grey Relation Analysis (GRA). Then, a new feature set based on residual clustering features is generated after the hierarchical clustering is applied to classify the characteristics of the residuals. Subsequently, Extreme Gradient Boosting (XGBoost) acts as the second layer that utilizes those residual clustering features to yield the prediction results. The final prediction is by incorporating the results from the first layer and second layer correspondingly. As for the performance evaluation, using agricultural product sales data from a supermarket in China from 1 July 2020 to 30 June 2023, the results demonstrate superiority over standalone RF and XGBoost, with a Mean Absolute Percentage Error (MAPE) reduction of 10% and 12%, respectively, and a coefficient of determination (R2) increase of 22% and 24%, respectively. Additionally, its generalization is validated across 42 types of agricultural products from six vegetable categories, showing its extensive practical ability. Such performances reveal that the proposed model beneficially enhances the precision of short-term agricultural product sales forecasting, with the advantages of optimizing the supply chain from producers to consumers and minimizing food waste accordingly.

Development of an artificial intelligent model for pre-endoscopic screening of precancerous lesions in gastric cancer

BackgroundGiven the high cost of endoscopy in gastric cancer (GC) screening, there is an urgent need to explore cost-effective methods for the large-scale prediction of precancerous lesions of gastric cancer (PLGC). We aim to construct a hierarchical artificial intelligence-based multimodal non-invasive method for pre-endoscopic risk screening, to provide tailored recommendations for endoscopy.MethodsFrom December 2022 to December 2023, a large-scale screening study was conducted in Fujian, China. Based on traditional Chinese medicine theory, we simultaneously collected tongue images and inquiry information from 1034 participants, considering the potential of these data for PLGC screening. Then, we introduced inquiry information for the first time, forming a multimodality artificial intelligence model to integrate tongue images and inquiry information for pre-endoscopic screening. Moreover, we validated this approach in another independent external validation cohort, comprising 143 participants from the China-Japan Friendship Hospital.ResultsA multimodality artificial intelligence-assisted pre-endoscopic screening model based on tongue images and inquiry information (AITonguequiry) was constructed, adopting a hierarchical prediction strategy, achieving tailored endoscopic recommendations. Validation analysis revealed that the area under the curve (AUC) values of AITonguequiry were 0.74 for overall PLGC (95% confidence interval (CI) 0.71–0.76, p < 0.05) and 0.82 for high-risk PLGC (95% CI 0.82–0.83, p < 0.05), which were significantly and robustly better than those of the independent use of either tongue images or inquiry information alone. In addition, AITonguequiry has superior performance compared to existing PLGC screening methodologies, with the AUC value enhancing 45% in terms of PLGC screening (0.74 vs. 0.51, p < 0.05) and 52% in terms of high-risk PLGC screening (0.82 vs. 0.54, p < 0.05). In the independent external verification, the AUC values were 0.69 for PLGC and 0.76 for high-risk PLGC.ConclusionOur AITonguequiry artificial intelligence model, for the first time, incorporates inquiry information and tongue images, leading to a higher precision and finer-grained pre-endoscopic screening of PLGC. This enhances patient screening efficiency and alleviates patient burden.

RMS-FlowNet++: Efficient and Robust Multi-scale Scene Flow Estimation for Large-Scale Point Clouds

The proposed RMS-FlowNet++ is a novel end-to-end learning-based architecture for accurate and efficient scene flow estimation that can operate on high-density point clouds. For hierarchical scene flow estimation, existing methods rely on expensive Farthest-Point-Sampling (FPS) to sample the scenes, must find large correspondence sets across the consecutive frames and/or must search for correspondences at a full input resolution. While this can improve the accuracy, it reduces the overall efficiency of these methods and limits their ability to handle large numbers of points due to memory requirements. In contrast to these methods, our architecture is based on an efficient design for hierarchical prediction of multi-scale scene flow. To this end, we develop a special flow embedding block that has two advantages over the current methods: First, a smaller correspondence set is used, and second, the use of Random-Sampling (RS) is possible. In addition, our architecture does not need to search for correspondences at a full input resolution. Exhibiting high accuracy, our RMS-FlowNet++ provides a faster prediction than state-of-the-art methods, avoids high memory requirements and enables efficient scene flow on dense point clouds of more than 250K points at once. Our comprehensive experiments verify the accuracy of RMS-FlowNet++ on the established FlyingThings3D data set with different point cloud densities and validate our design choices. Furthermore, we demonstrate that our model has a competitive ability to generalize to the real-world scenes of the KITTI data set without fine-tuning.

A new soft clustering method for traffic prediction in bike-sharing systems

For the efficient management of bike-sharing systems (BSSs), accurate demand predictions are crucial to address the uneven distribution of bikes at various stations. Recent studies have explored a hierarchical prediction framework using cluster-level models to more accurately estimate demand at the station level. However, in frameworks based on hard clustering, where each station is exclusively assigned to one of several clusters, prediction accuracy tends to be lower for stations at the cluster boundaries. To improve accuracy for such stations, this study proposes a novel soft clustering algorithm for BSSs. The key idea is to allow stations to belong to multiple clusters, calculating the membership degree for each station based on transitions between stations and clusters obtained through hard clustering. This study also investigated the impact of restricting clusters to which individual stations belong based on distance or usage history. Two approaches, distance- and usage-based, were employed to determine the clusters to which each station belongs. Experimental results using Seoul Bike data demonstrate the effectiveness of the proposed method in enhancing traffic prediction accuracy within the hierarchical prediction framework. Notably, excluding clusters with minimal usage for each station using the usage-based approach yielded the best performance.

Hierarchical Predictions of Fine-to-Coarse Time Span and Atmospheric Field Reconstruction for Typhoon Track Prediction

The prediction of typhoon tracks in the Northwest Pacific is key to reducing human casualties and property damage. Traditional numerical forecasting models often require substantial computational resources, are high-cost, and have significant limitations in prediction speed. This research is dedicated to using deep learning methods to address the shortcomings of traditional methods. Our method (AFR-SimVP) is based on a large-kernel convolutional spatio-temporal prediction network combined with multi-feature fusion for forecasting typhoon tracks in the Northwest Pacific. In order to more effectively suppress the effect of noise in the dataset to enhance the generalization ability of the model, we use a multi-branch structure, incorporate an atmospheric reconstruction subtask, and propose a second-order smoothing loss to further improve the prediction ability of the model. More importantly, we innovatively propose a multi-time-step typhoon prediction network (HTAFR-SimVP) that does not use the traditional recurrent neural network family of models at all. Instead, through fine-to-coarse hierarchical temporal feature extraction and dynamic self-distillation, multi-time-step prediction is achieved using only a single regression network. In addition, combined with atmospheric field reconstruction, the network achieves integrated prediction for multiple tasks, which greatly enhances the model’s range of applications. Experiments show that our proposed network achieves optimal performance in the 24 h typhoon track prediction task. Our regression network outperforms previous recurrent network-based typhoon prediction models in the multi-time-step prediction task and also performs well in multiple integration tasks.

Hierarchical prediction of dam deformation based on hybrid temporal network and load-oriented residual correction

Dam deformation prediction methods try to make the predicted values infinitely close to the true values, but no method can capture all the deformation information and residuals are inevitable. For this reason, this paper develops a hierarchical dam deformation prediction method centred on residual correction strategy combined with hybrid temporal network. Firstly, hybrid CNN-LSTM is established based on convolutional neural network (CNN) and long short-term memory network (LSTM), aiming at mining the deformation temporal features and outputting the deformation trend information. Secondly, in the training stage, the mapping relationship between environmental loads and residuals is established by using Extreme Learning Machine (ELM), and the load information embedded in the residuals is mined to output the fluctuation information specific to dam deformation. Then, in the prediction stage, the preliminary deformation prediction results are outputted by CNN-LSTM, and the residual correction values are obtained according to the mapping relationship between environmental loads and residuals, so as to obtain the prediction results. The analysis shows that the outputs of CNN-LSTM can reflect the deformation trend term and provide reliable preliminary prediction results for residual correction, the ELM mapping relation model can capture the loading information embedded in the residuals that cannot be mined by the temporal model, which improves the interpretability of residual correction, and the output results reflect the deformation fluctuations well. The proposed method can realise the adaptive correction of the preliminary deformation prediction results, and the prediction accuracy is satisfactory and adaptable.

MSFA: Multi‐stage feature aggregation network for multi‐label image recognition

AbstractMulti‐label image recognition (MLR) is a significant branch of image classification that aims to assign multiple categorical labels to each input. Previous research has focused on enhancing the learning of category‐related regional features. However, the potential impact of multi‐scale distributions in intra‐ and inter‐category targets on MLR tends to be neglected. Besides, semantic consistency for categories is restricted to be considered on single‐scale features, resulting in suboptimal feature extraction. To address the limitations of above, a Multi‐stage Feature Aggregation (MSFA) network is proposed. In MSFA, a novel local feature extraction method is suggested to progressively extract category‐related high‐resolution local features in both spatial and channel dimensions. Subsequently, local and global features are fused without additional up‐ and down‐sampling to enrich the scale diversity of the features while incorporating refined class‐specific information. Furthermore, a hierarchical prediction scheme for MLR is proposed, which generates classification confidence corresponding to different scales under hierarchical loss supervision. Consequently, the final output of the network comes from the joint prediction by the classifiers on multi‐scale features, ensuring a stronger feature extraction capability. The extensive experiments have been carried on VOC and MS‐COCO datasets, and the superiority of MSFA over existing mainstream methods has been verified.

Light field image coding using a residual channel attention network–based view synthesis

PurposeLight field images (LFIs) have gained popularity as a technology to increase the field of view (FoV) of plenoptic cameras since they can capture information about light rays with a large FoV. Wide FoV causes light field (LF) data to increase rapidly, which restricts the use of LF imaging in image processing, visual analysis and user interface. Effective LFI coding methods become of paramount importance. This paper aims to eliminate more redundancy by exploring sparsity and correlation in the angular domain of LFIs, as well as mitigate the loss of perceptual quality of LFIs caused by encoding.Design/methodology/approachThis work proposes a new efficient LF coding framework. On the coding side, a new sampling scheme and a hierarchical prediction structure are used to eliminate redundancy in the LFI's angular and spatial domains. At the decoding side, high-quality dense LF is reconstructed using a view synthesis method based on the residual channel attention network (RCAN).FindingsIn three different LF datasets, our proposed coding framework not only reduces the transmitted bit rate but also maintains a higher view quality than the current more advanced methods.Originality/value(1) A new sampling scheme is designed to synthesize high-quality LFIs while better ensuring LF angular domain sparsity. (2) To further eliminate redundancy in the spatial domain, new ranking schemes and hierarchical prediction structures are designed. (3) A synthetic network based on RCAN and a novel loss function is designed to mitigate the perceptual quality loss due to the coding process.

A Hierarchical Prediction Method for Pedestrian Head Injury in Intelligent Vehicle with Combined Active and Passive Safety System.

With the development of intelligent vehicle technology, the probability of road traffic accidents occurring has been effectively reduced to a certain extent. However, there is still insufficient research on head injuries in human vehicle collisions, making it impossible to effectively predict pedestrian head injuries in accidents. To study the efficacy of a combined active and passive safety system on pedestrian head protection through the combined effect of the exterior airbag and the braking control systems of an intelligent vehicle, a "vehicle-pedestrian" interaction system is constructed in this study and is verified by real collision cases. On this basis, a combined active and passive system database is developed to analyze the cross-influence of the engine hood airbag and the vehicle braking curve parameters on pedestrian HIC (head injury criterion). Meanwhile, a hierarchy design strategy for a combined active and passive system is proposed, and a rapid prediction of HIC is achieved via the establishment of a fitting equation for each grading. The results show that the exterior airbag can effectively protect the pedestrian's head, prevent the collision between the pedestrian's head and the vehicle front structure, and reduce the HIC. The braking parameter H2 is significantly correlated with head injury, and when H2 is less than 1.8, the HIC value is less than 1000 in nearly 90% of cases. The hierarchy design strategy and HIC prediction method of the combined active and passive system proposed in this paper can provide a theoretical basis for rapid selection and parameter design.

A comprehensive patient-specific prediction model for temporomandibular joint osteoarthritis progression

Temporomandibular joint osteoarthritis (TMJ OA) is a prevalent degenerative disease characterized by chronic pain and impaired jaw function. The complexity of TMJ OA has hindered the development of prognostic tools, posing a significant challenge in timely, patient-specific management. Addressing this gap, our research employs a comprehensive, multidimensional approach to advance TMJ OA prognostication. We conducted a prospective study with 106 subjects, 74 of whom were followed up after 2 to 3 y of conservative treatment. Central to our methodology is the development of an innovative, open-source predictive modeling framework, the Ensemble via Hierarchical Predictions through Nested cross-validation tool (EHPN). This framework synergistically integrates 18 feature selection, statistical, and machine learning methods to yield an accuracy of 0.87, with an area under the ROC curve of 0.72 and an F1 score of 0.82. Our study, beyond technical advancements, emphasizes the global impact of TMJ OA, recognizing its unique demographic occurrence. We highlight key factors influencing TMJ OA progression. Using SHAP analysis, we identified personalized prognostic predictors: lower values of headache, lower back pain, restless sleep, condyle high gray level-GL-run emphasis, articular fossa GL nonuniformity, and long-run low GL emphasis; and higher values of superior joint space, mouth opening, saliva Vascular-endothelium-growth-factor, Matrix-metalloproteinase-7, serum Epithelial-neutrophil-activating-peptide, and age indicate recovery likelihood. Our multidimensional and multimodal EHPN tool enhances clinicians' decision-making, offering a transformative translational infrastructure. The EHPN model stands as a significant contribution to precision medicine, offering a paradigm shift in the management of temporomandibular disorders and potentially influencing broader applications in personalized healthcare.

SHAPE: A Sample-Adaptive Hierarchical Prediction Network for Medication Recommendation.

Effectively medication recommendation with complex multimorbidity conditions is a critical yet challenging task in healthcare. Most existing works predicted medications based on longitudinal records, which assumed the encoding format of intra-visit medical events are serialized and information transmitted patterns of learning longitudinal sequence data are stable. However, the following conditions may have been ignored: 1) A more compact encoder for intra-relationship in the intra-visit medical event is urgent; 2) Strategies for learning accurate representations of the variable longitudinal sequences of patients are different. In this article, we proposed a novel Sample-adaptive Hierarchical medicAtion Prediction nEtwork, termed SHAPE, to tackle the above challenges in the medication recommendation task. Specifically, we design a compact intra-visit set encoder to encode the relationship in the medical event for obtaining visit-level representation and then develop an inter-visit longitudinal encoder to learn the patient-level longitudinal representation efficiently. To endow the model with the capability of modeling the variable visit length, we introduce a soft curriculum learning method to assign the difficulty of each sample automatically by the visit length. Extensive experiments on a benchmark dataset verify the superiority of our model compared with several state-of-the-art baselines.

Spatio-temporal layers based intra-operative stereo depth estimation network via hierarchical prediction and progressive training

Background and ObjectiveSafety of robotic surgery can be enhanced through augmented vision or artificial constraints to the robotl motion, and intra-operative depth estimation is the cornerstone of these applications because it provides precise position information of surgical scenes in 3D space. High-quality depth estimation of endoscopic scenes has been a valuable issue, and the development of deep learning provides more possibility and potential to address this issue. MethodsIn this paper, a deep learning-based approach is proposed to recover 3D information of intra-operative scenes. To this aim, a fully 3D encoder-decoder network integrating spatio-temporal layers is designed, and it adopts hierarchical prediction and progressive learning to enhance prediction accuracy and shorten training time. ResultsOur network gets the depth estimation accuracy of MAE 2.55±1.51 (mm) and RMSE 5.23±1.40 (mm) using 8 surgical videos with a resolution of 1280×1024, which performs better compared with six other state-of-the-art methods that were trained on the same data. ConclusionsOur network can implement a promising depth estimation performance in intra-operative scenes using stereo images, allowing the integration in robot-assisted surgery to enhance safety.

The Design Principle of Allostatic Mechanisms: The Free Energy Principle

Homeostasis is maintained by neural reflex controls, which attempt to maintain the body's internal organs within a predetermined range. However, when deviations are significant, homeostasis may not be restored by these standard controls. Therefore, a function called allostasis is necessary to maximize the energy efficiency of bodily functions, anticipate changing needs (e.g., a future increase in body temperature or oxygen demand), and regulate the body's state before dyshomeostasis occurs. In terms of the free energy principle (FEP), the expectation to survive (i.e., to stay within established confines and to maintain homeostatic integrity over time) is realized by a hierarchical prediction mechanism. Allostasis requires the integration of exteroceptive, interoceptive, and proprioceptive signals. In this paper, the neural mechanism of allostasis is discussed in terms of the FEP.

Inter- and intra-uncertainty based feature aggregation model for semi-supervised histopathology image segmentation

Acquiring pixel-level annotations is often limited in applications such as histology studies that require domain expertise. Various semi-supervised learning approaches have been developed to work with limited ground truth annotations, such as the popular teacher-student models. However, hierarchical prediction uncertainty within the student model (intra-uncertainty) and image prediction uncertainty (inter-uncertainty) have not been fully utilized by existing methods. To address these issues, we first propose a novel inter- and intra-uncertainty regularization method to measure and constrain both inter- and intra-inconsistencies in the teacher-student architecture. We also propose a new two-stage network with pseudo-mask guided feature aggregation (PG-FANet) as the segmentation model. The two-stage structure complements with the uncertainty regularization strategy to avoid introducing extra modules in solving uncertainties and the aggregation mechanisms enable multi-scale and multi-stage feature integration. Comprehensive experimental results over the MoNuSeg and CRAG datasets show that our PG-FANet outperforms other state-of-the-art methods and our semi-supervised learning framework yields competitive performance with a limited amount of labeled data.

Two-stage Hierarchical Framework for Solar Flare Prediction

Solar flares, often accompanied by coronal mass ejections and other solar phenomena, are one of the most important sources affecting space weather. It is important to investigate the forecast approach of solar flares to mitigate their destructive effect on the Earth. Statistical analysis, associated with data from 2010 to 2017 in Space-weather HMI Active Region Patches (SHARPs) collected by the Solar Dynamics Observatory's Helioseismic and Magnetic Imager, reveals that there is a distribution divergence between the two types of active regions (ARs) of solar flares. A two-stage hierarchical prediction framework is formulated to better utilize this intrinsic distribution information. Specially, we pick up the ARs where at least one solar flare event occurs within the next 48 hr as flaring ARs through balanced random forest and naive Bayesian methods and then predict the events from flaring ARs by a cascade module of learning models. The empirical evaluation of SHARPs data from 2016 to 2019 verifies the promising performance of our framework, e.g., 0.727 for the true skill statistic.

Modeling the Wintering Habitat Distribution of the Black Stork in Shaanxi, China: A Hierarchical Integration of Climate and Land Use/Land Cover Data.

Species distribution models (SDMs) are effective tools for wildlife conservation and management, as they employ the quantification of habitat suitability and environmental niches to evaluate the patterns of species distribution. The utilization of SDMs at various scales in a hierarchical approach can provide additional and complementary information, significantly improving decision-making in local wildlife conservation initiatives. In this study, we considered the appropriate spatial scale and data resolution to execute species distribution modeling, as these factors greatly influence the modeling procedures. We developed SDMs for wintering black storks at both the regional and local scales. At the regional scale, we used climatic and climate-driven land use/land cover (LULC) variables, along with wintering occurrence points, to develop models for mainland China. At the local scale, we used local environmental variables and locally gathered wintering site data to develop models for Shaanxi province. The predictions from both the regional and local models were then combined at the provincial level by overlapping suitable areas based on climatic and local conditions. We compared and evaluated the resulting predictions using seven statistical metrics. The national models provide information on the appropriate climatic conditions for the black stork during the wintering period throughout China, while the provincial SDMs capture the important local ecological factors that influence the suitability of habitats at a finer scale. As anticipated, the national SDMs predict a larger extent of suitable areas compared to the provincial SDMs. The hierarchical prediction approach is considered trustworthy and, on average, yields better outcomes than non-hierarchical methods. Our findings indicate that human-driven LULC changes have a significant and immediate impact on the wintering habitat of the black stork. However, the effects of climate change seem to be reducing the severity of this impact. The majority of suitable wintering habitats lie outside the boundaries of protected areas, highlighting the need for future conservation and management efforts to prioritize addressing these conservation gaps and focusing on the protection of climate refuges.

Aberrant Hierarchical Prediction Errors Are Associated With Transition to Psychosis: A Computational Single-Trial Analysis of the Mismatch Negativity

BackgroundMismatch negativity reductions are among the most reliable biomarkers for schizophrenia and have been associated with increased risk for conversion to psychosis in individuals who are at clinical high risk for psychosis (CHR-P). Here, we adopted a computational approach to develop a mechanistic model of mismatch negativity reductions in CHR-P individuals and patients early in the course of schizophrenia. MethodsElectroencephalography was recorded in 38 CHR-P individuals (15 converters), 19 patients early in the course of schizophrenia (≤5 years), and 44 healthy control participants during three different auditory oddball mismatch negativity paradigms including 10% duration, frequency, or double deviants, respectively. We modeled sensory learning with the hierarchical Gaussian filter and extracted precision-weighted prediction error trajectories from the model to assess how the expression of hierarchical prediction errors modulated electroencephalography amplitudes over sensor space and time. ResultsBoth low-level sensory and high-level volatility precision-weighted prediction errors were altered in CHR-P individuals and patients early in the course of schizophrenia compared with healthy control participants. Moreover, low-level precision-weighted prediction errors were significantly different in CHR-P individuals who later converted to psychosis compared with nonconverters. ConclusionsOur results implicate altered processing of hierarchical prediction errors as a computational mechanism in early psychosis consistent with predictive coding accounts of psychosis. This computational model seems to capture pathophysiological mechanisms that are relevant to early psychosis and the risk for future psychosis in CHR-P individuals and may serve as predictive biomarkers and mechanistic targets for the development of novel treatments.

Benefits of hierarchical predictions for digital soil mapping—An approach to map bimodal soil pH

Maps of soil pH are an important tool for making decisions in sustainable forest management. Accurate pH mapping, therefore, is crucial to support decisions by authorities or forest companies. Soil pH values typically exhibit a distinct distribution characterized by two frequently encountered pH ranges, wherein aluminium oxides (Al2O3) and carbonates (CaCO3) act as the primary buffer agents. Soil samples with moderately acid pH values (pH CaCl2 of 4.5-6) are less commonly observed due to their weaker buffering capacity. The different strength of buffer agents results in a distinct bimodal distribution of soil pH values with peaks at pH of around 4 and 7.5. Commonly used approaches for spatial mapping neglect this often observed characteristic of soil pH and predict unimodal distributions with too many moderately acid pH values. For ecological map applications this might result in misleading interpretations.This article presents a novel approach to produce pH maps that are able to reproduce pedogenic processes. The procedure is suitable for bimodal responses where the response distribution is naturally inherent and needs to be reproduced for the predictions. It is model-agnostic, namely independent from the used statistical prediction method. Calibration data is optimally split into two parts corresponding each to a data culmination, i.e. for soil pH values belonging to the ranges of the two principal buffer agents (Al2O3 and CaCO3). For each subset a separate model is then built. In addition, a binary model is fitted to assign every new prediction location a probability to belong either to Al2O3 or CaCO3 buffer range. Predictions are combined by weighted mean. Weights are derived from probabilities predicted by the binary model. Degree of smoothness is chosen by sigmoid transform which allows for optimal continuous transition of the pH values between Al2O3 and CaCO3 buffer ranges. For each location uncertainty distributions may be combined by using the same weights.We illustrated application of the new approach to a medium and strong bimodal distributed response (1) pH in 0–5 cm and (2) pH in 60–100 cm of forest soils in Switzerland (2530 calibration sites). While model performance measured at 354 validation sites slightly dropped compared to a common modelling approach (drop of R2 of 0.02–0.03) distributional properties of the predictions are much more meaningful from a pedogenic point of view. We were able to demonstrate the benefits of considering specific distributional properties of responses within the prediction process and expanded model assessment by comparing observed and predicted distributions.

MOHP-EC: A Multiobjective Hierarchical Prediction Framework for Urban Rail Transit Passenger Flow

For sophisticated management, advertisement placement, and epidemic prevention control of urban rail transit (URT), accurate and real-time predictions of passenger flows at different levels are of great importance. Unlike traditional prediction tasks, hierarchical prediction (HP) requires that the hierarchical constraints be satisfied as much as possible (the sum of the predicted passenger flow of child nodes should nearly equal the parent node) to achieve realistic predictions. This article proposes a multiobjective HP (MOHP) framework with an error compensation (EC) mechanism for predicting URT passenger flow with a hierarchical structure. Three components are included: the initial prediction module, the EC module, and the hierarchical coordination module. In the initial prediction module, the initial passenger flow prediction of each layer is carried out. The EC model is developed based on proportional-integral-derivative control to compensate for the initial predicted value of every layer. As a final step, a trainable HP model is constructed based on deep learning to coordinate the prediction values of each layer. As examples, we construct three scenarios of passenger flow hierarchy based on the URT system in Wuxi, China. The constructed prediction framework is used to conduct experimental analyses. As a result, the MOHP-EC prediction framework could satisfy the hierarchical constraints and use the passenger flow hierarchical information to reduce prediction errors. The mean absolute error was reduced by 35%, and the root-mean-square error was reduced by 39%.

Hierarchical transfer learning with applications to electricity load forecasting

The recent abundance of electricity consumption data available at different scales provides new opportunities and highlights the need for new techniques to leverage information present at finer scales in order to improve forecasts at wider scales. In this study, we take advantage of the similarity between this hierarchical prediction problem and transfer learning where source data are observed at a low aggregation level and target data at a global level. We develop two methods for hierarchical transfer learning based on stacking generalized additive models and random forests (GAM-RF). We also propose and compare adaptations of online aggregation of experts in a hierarchical context using quantile GAM-RF as experts. We apply these methods to two electricity load forecasting problems at the national scale by using smart meter data in the first case and regional data in the second case. For these two user cases, we compared the performance of our methods and benchmark algorithms, and investigated their behavior using variable importance analysis. Our results demonstrate that both methods can lead to significantly improved predictions.