Stacking Strategy Research Articles

Pseudo label estimation based on label distribution optimization for industrial semi-supervised soft sensor

In process industry, the lack of sufficient labeled data often leads to poor performance of traditional supervised soft sensors. Thus, a pseudo label estimation method based on label distribution optimization (PLELDO) is proposed. PLELDO first converts the pseudo label estimation into an explicit label distribution optimization problem, and then obtains high-confidence pseudo labels. Further, a semi-supervised ensemble soft sensor modeling framework namely EnPLELDO is developed. EnPLELDO first obtains sufficient good pseudo labeled data by repeating small-scale pseudo label estimation. Then, a set of diverse base models are constructed from the extended training sets. Finally, these enhanced base models are combined through stacking strategy. The application results from an industrial fed-batch fermentation process show that, compared with several state-of-the-art methods, PLELDO can obtain better pseudo label estimations and EnPLELDO can deliver more accurate predictions of key quality variables by leveraging the scarce labeled data and large-scale unlabeled data.

Additive manufactured Triply Periodical Minimal Surface lattice structures with modulated hybrid topology

The fabrication freedom offered by additive manufacturing techniques is a unique asset to be exploited in the design of lightweight lattice structures. Adjusting topological and architectural features towards the design of multi-morphological lattice structures can offer a high potential towards obtaining fine-tuned mechanical response. Herein, to understand the role of unit cell topology and arrangement, various stacking and gradient strategies were implemented to modulate the overall mechanical response of the lattice structures under compression, using two Triply Periodical Minimal Surface (TPMS) unit cell designs. Experimental and numerical approaches were developed to reveal the deformation mechanism and failure modes and quantify stiffness, quasi-static uniaxial compressive strength, and energy absorption capacity of the structures. The topological arrangement of the selected unit cells was found to play a key role in defining the mechanical performance of the designed lattice structures. The obtained results demonstrated the high potential of various graded design elements for obtaining lattice structures with desired properties.

Fabrication and characterizations of Zn-doped SnO2-Ti4O7 anode for electrochemical degradation of hexafluoropropylene oxide dimer acid and its homologues

Hexafluoropropylene oxide dimer acid (HFPO-DA) and its homologues, as perfluorinated ether alkyl substances with strong antioxidant properties, have rarely been reported by electrooxidation processes to achieve good results. Herein, we report the use of an oxygen defect stacking strategy to construct Zn-doped SnO2-Ti4O7 for the first time and enhance the electrochemical activity of Ti4O7. Compared with the original Ti4O7, the Zn-doped SnO2-Ti4O7 showed a 64.4% reduction in interfacial charge transfer resistance, a 17.5% increase in the cumulative rate of •OH generation, and an enhanced oxygen vacancy concentration. The Zn-doped SnO2-Ti4O7 anode exhibited high catalytic efficiency of 96.4% for HFPO-DA within 3.5 h at 40 mA/cm2. Hexafluoropropylene oxide trimer and tetramer acid exhibit more difficult degradation due to the protective effect of the -CF3 branched chain and the addition of the ether oxygen atom leading to a significant increase in the C-F bond dissociation energy. The degradation rates of 10 cyclic degradation experiments and the leaching concentrations of Zn and Sn after 22 electrolysis experiments demonstrated the good stability of the electrodes. In addition, the aqueous toxicity of HFPO-DA and its degradation products was evaluated. This study analyzed the electrooxidation process of HFPO-DA and its homologues for the first time, and provided some new insights.

Heating choices and residential willingness to pay for clean heating: Evidence from a household survey in rural China

Due to the severity of climate change, the Chinese government has committed to the carbon neutrality target. Winter heating is a major source of carbon emissions in northern China, and it is urgent to promote the adoption of clean heating in northern rural areas. With indoor-survey data, both two-stage regression and structural equation modeling were used in this study to figure out the exact driving factors and causal relationships of rural households' energy choices considering subjective factors that may influence the energy choices and willingness to pay. Results contradict the energy ladder hypothesis and conclude that it might be caused by energy consumption habits and fuel stacking strategy. Both demographic characteristics and education level have a significant effect on heating energy choices. As for residents' willingness to pay for clean heating, it shows that households’ environmental perceptions, socioeconomic status, and demand for policy subsidies play an important role. Our results make a strong case for policies that shape residents' low-carbon values, promote technological innovation in clean energy, and set reasonable subsidies for clean heating to encourage a clean heating transition.

An evolutionary ensemble model based on GA for epidemic transmission prediction

This paper proposes an evolutionary ensemble model based on a Genetic Algorithm (GAEEM) to predict the transmission trend of infectious diseases based on ensemble again and prediction again. The model utilizes the strong global optimization capability of GA for tuning the ensemble structure. Compared with the traditional ensemble learning model, GAEEM has three main advantages: 1) It is set to address the problems of information leakage in the traditional Stacking strategy and overfitting in the Blending strategy. 2) It uses a GA to optimize the combination of base learners and determine the sub. 3) The feature dimension of the data used in this layer is extended based on the optimal base learner combination prediction information data, which can reduce the risk of underfitting and increase prediction accuracy. The experimental results show that the R2 performance of the model in the six cities data set is higher than all the comparison models by 0.18 on average. The MAE and MSE are lower than 42.98 and 42,689.72 on average. The fitting performance is more stable in each data set and shows good generalization, which can predict the epidemic spread trend of each city more accurately.

Dynamic Mechanical Strength Prediction of BFRC Based on Stacking Ensemble Learning and Genetic Algorithm Optimization

Split Hopkinson pressure bar (SHPB) tests are usually used to determine the dynamic mechanical strength of basalt-fiber-reinforced concrete (BFRC), but this test method is time-consuming and expensive. This paper makes predictions about the dynamic mechanical strength of BFRC by employing machine learning (ML) algorithms and feature sets drawn from experimental data from prior works. However, there is still the problem of improving the accuracy of the dynamic mechanical strength prediction by the BFRC, which remains a challenge. Using stacking ensemble learning and genetic algorithms (GA) to optimize parameters, this study proposes a prediction method that combines these two techniques for obtaining accurate predictions. This method is composed of three parts: (1) the training uses multiple base learners, and the algorithms employed by the learners include extreme gradient boosting (XGBoost), gradient boosting (GB), random forest (RF), and support vector regression (SVR); (2) multi-base learners are combined using a stacking strategy to obtain the final prediction; and (3) using GA, the parameters are optimized in the prediction model. An experiment was conducted to compare the proposed approach with popular techniques for machine learning. In the study, the stacking ensemble algorithm integrated the base learner prediction results to improve the model’s performance and the GA further improved prediction accuracy. As a result of the application of the method, the dynamic mechanical strength of BFRC can be predicted with high accuracy. A SHAP analysis was also conducted using the stacking model to determine how important the contributing properties are and the sensitivity of the stacking model. Based on the results of this study, it was found that in the SHPB test, the strain rate had the most significant influence on the DIF, followed by the specimen diameter and the compressive strength.

Ensemble of 6 DoF Pose estimation from state-of-the-art deep methods.

Deep learning methods have revolutionized computer vision since the appearance of AlexNet in 2012. Nevertheless, 6 degrees of freedom pose estimation is still a difficult task to perform precisely. Therefore, we propose 2 ensemble techniques to refine poses from different deep learning 6DoF pose estimation models. The first technique, merge ensemble, combines the outputs of the base models geometrically. In the second, stacked generalization, a machine learning model is trained using the outputs of the base models and outputs the refined pose. The merge method improves the performance of the base models on LMO and YCB-V datasets and performs better on the pose estimation task than the stacking strategy.

Quantitative Assessment of Apple Mosaic Disease Severity Based on Hyperspectral Images and Chlorophyll Content

The infection of Apple mosaic virus (ApMV) can severely damage the cellular structure of apple leaves, leading to a decrease in leaf chlorophyll content (LCC) and reduced fruit yield. In this study, we propose a novel method that utilizes hyperspectral imaging (HSI) technology to non-destructively monitor ApMV-infected apple leaves and predict LCC as a quantitative indicator of disease severity. LCC data were collected from 360 ApMV-infected leaves, and optimal wavelengths were selected using competitive adaptive reweighted sampling algorithms. A high-precision LCC inversion model was constructed based on Boosting and Stacking strategies, with a validation set Rv2 of 0.9644, outperforming traditional ensemble learning models. The model was used to invert the LCC distribution image and calculate the average and coefficient of variation (CV) of LCC for each leaf. Our findings indicate that the average and CV of LCC were highly correlated with disease severity, and their combination with sensitive wavelengths enabled the accurate identification of disease severity (validation set overall accuracy = 98.89%). Our approach considers the role of plant chemical composition and provides a comprehensive evaluation of disease severity at the leaf scale. Overall, our study presents an effective way to monitor and evaluate the health status of apple leaves, offering a quantifiable index of disease severity that can aid in disease prevention and control.

Multifunctional Soft Stackable Robots by Netting-Rolling-Splicing Pneumatic Artificial Muscles.

Soft robots equipped with multifunctionalities have been increasingly needed for secure, adaptive, and autonomous functioning in unknown and unpredictable environments. Robotic stacking is a promising solution to increase the functional diversity of soft robots, which are required for safe human-machine interactions and adapting in unstructured environments. However, most existing multifunctional soft robots have a limited number of functions or have not fully shown the superiority of the robotic stacking method. In this study, we present a novel robotic stacking strategy, Netting-Rolling-Splicing (NRS) stacking, based on a dimensional raising method via 2D-to-3D rolling-and-splicing of netted stackable pneumatic artificial muscles to quickly and efficiently fabricate multifunctional soft robots based on the same, simple, and cost-effective elements. To demonstrate it, we developed a TriUnit robot that can crawl 0.46 ± 0.022 body length per second (BL/s) and climb 0.11 BL/s, and can carry a 3 kg payload while climbing. Also, the TriUnit can be used to achieve novel omnidirectional pipe climbing including rotating climbing, and conduct bionic swallowing-and-regurgitating, multi-degree-of-freedom manipulation based on their multimodal combinations. Apart from these, steady rolling, with a speed of 0.19 BL/s, can be achieved by using a pentagon unit. Furthermore, we applied the TriUnit pipe climbing robot in panoramic shooting and cargo transferring to demonstrate the robot's adaptability for different tasks. The NRS stacking-driven soft robot here has demonstrated the best overall performance among existing stackable soft robots, representing a new and effective way for building multifunctional and multimodal soft robots in a cost-effective and efficient way.

A Diabetes Prediction System Based on Incomplete Fused Data Sources

In recent years, the diabetes population has grown younger. Therefore, it has become a key problem to make a timely and effective prediction of diabetes, especially given a single data source. Meanwhile, there are many data sources of diabetes patients collected around the world, and it is extremely important to integrate these heterogeneous data sources to accurately predict diabetes. For the different data sources used to predict diabetes, the predictors may be different. In other words, some special features exist only in certain data sources, which leads to the problem of missing values. Considering the uncertainty of the missing values within the fused dataset, multiple imputation and a method based on graph representation is used to impute the missing values within the fused dataset. The logistic regression model and stacking strategy are applied for diabetes training and prediction on the fused dataset. It is proved that the idea of combining heterogeneous datasets and imputing the missing values produced in the fusion process can effectively improve the performance of diabetes prediction. In addition, the proposed diabetes prediction method can be further extended to any scenarios where heterogeneous datasets with the same label types and different feature attributes exist.

Storage strategy of outbound containers with uncertain weight by data-driven hybrid genetic simulated annealing algorithm.

It is necessary to ensure the ship's stability in container ship stowage and loading and unloading containers. This work aims to reduce the container dumping operation at the midway port and improve the efficiency of ship transportation. Firstly, the constraint problem of the traditional container ship stacking is introduced to realize the multi-condition mathematical model of the container ship, container, and wharf. Secondly, a Hybrid Genetic and Simulated Annealing Algorithm (HGSAA) model is proposed for the container stacking and loading stacking in the yard. The specific container space allocation and multi-yard crane adjustment scheme are studied. Finally, the effectiveness of the multi-condition container ship stowage model is verified by numerical experiments by changing the number of outbound containers, storage strategies, storage yards, and bridges. The experimental results show that the HGSAA mode converges to 106.1min at the 751st iteration. Of these, the non-loading and unloading time of yard bridge 1 is 3.43min. The number of operating boxes is 25. The non-loading and unloading time of yard bridge 2 is 3.2min, and the operating box volume is 25 boxes. The objective function of the genetic algorithm converges when it iterates to generation 903 and 107.9min. Among them, the non-loading and unloading time of yard bridge 1 is 4.1min. The non-loading and unloading time of yard bridge 2 is 3.1min. Therefore, the proposed HGSAA has a faster convergence speed than the genetic algorithm and can obtain relatively good results. The proposed container stacking strategy can effectively solve the specific container allocation and multi-yard crane scheduling problems. The finding provides a reference for optimizing container scheduling and improving shipping transportation efficiency.

Heterogeneous stacking strategy towards carbon aerogel for thermal management and electromagnetic interference shielding

Faced with the increasing heat dissipation and electromagnetic interference (EMI) shielding problems in electronics, carbon modified polymer-based composites with significant EMI shielding and thermal management performance are of particular interest. Herein, we proposed a carbon heterogeneous stacking strategy to construct the all carbon aerogels for the modification of epoxy resin. A hybrid 3D structure of reduced graphene oxide–carbon nanotube-vertical edge-rich graphene (rGO-CNT-VG) with covalent bonding are achieved all by chemical vapor deposition, in which the pore space of skeleton is creatively modified. Due to the elaborate design and control of microstructures, the obtained hierarchical 3D rGO-CNT-VG skeleton have plenty of seamlessly bonded heterogeneous interfaces between different components, which can create additional charge polarization, interfacial polarization and dielectric relaxation to promote significantly electromagnetic microwave attenuation and conversion and achieve ideal EMI shielding performance. Impressively, the rGO-CNT-VG/epoxy composites possess excellent thermal conductivity of 2.46 W m−1 K−1 and EMI shielding effectiveness of 56.65 dB, which are 5.1 times and 1.9 times higher than those of the rGO/epoxy composites, respectively. More importantly, the strategy of designing all carbon heterogeneous stacking skeleton in this study provides a guidance for synergistic controlling of multifunctional performance of composites.

Bulk Solids Stacking Strategy of a Rectangular Ship Cabin

Stacking of bulk solids in terminals, ships, trains, and other bulk solids storage yards is always challenging, considering the requirements of optimal utilization of storage areas. In this study, stacking shapes of a variety of bulk solids with different particle sizes were tested, and the curved shapes of the stockpiles were extracted to establish the actual three-dimensional models of the bulk solids accordingly. The three-dimensional curved stockpiles were used to design the bulk solids stacking strategy where the bulk stacking locations, the stacking volume, and the stacking flatness were optimized. A modified golden section method with a self-influenced factor was developed to improve the calculation efficiency of the stacking algorithm for bulk solids stacking flatness. Furthermore, the bulk solids stacking strategy and algorithm were verified by experiments. The results showed that the curved three-dimensional models were very close to the actual shapes of the bulk solids stockpiles, while the improved golden section method was more accurate and efficient than the traditional golden section method in determining the optimal values of the stacking volumes in terms of flatness. For the different bulk solids tested, the experiment results showed that good flat stacking can be achieved by using the developed stacking strategy.

M6A-BERT-Stacking: A Tissue-Specific Predictor for Identifying RNA N6-Methyladenosine Sites Based on BERT and Stacking Strategy

As the most abundant RNA methylation modification, N6-methyladenosine (m6A) could regulate asymmetric and symmetric division of hematopoietic stem cells and play an important role in various diseases. Therefore, the precise identification of m6A sites around the genomes of different species is a critical step to further revealing their biological functions and influence on these diseases. However, the traditional wet-lab experimental methods for identifying m6A sites are often laborious and expensive. In this study, we proposed an ensemble deep learning model called m6A-BERT-Stacking, a powerful predictor for the detection of m6A sites in various tissues of three species. First, we utilized two encoding methods, i.e., di ribonucleotide index of RNA (DiNUCindex_RNA) and k-mer word segmentation, to extract RNA sequence features. Second, two encoding matrices together with the original sequences were respectively input into three different deep learning models in parallel to train three sub-models, namely residual networks with convolutional block attention module (Resnet-CBAM), bidirectional long short-term memory with attention (BiLSTM-Attention), and pre-trained bidirectional encoder representations from transformers model for DNA-language (DNABERT). Finally, the outputs of all sub-models were ensembled based on the stacking strategy to obtain the final prediction of m6A sites through the fully connected layer. The experimental results demonstrated that m6A-BERT-Stacking outperformed most of the existing methods based on the same independent datasets.

A forecasting method of multi-category product sales: analysis and application

To solve the problems of high prediction costs and difficult practices in multi-category product classification in the retail industry, optimize the inventory, and improve resilience, this work introduces a forecasting method for multi-category product sales. The forecasting method divides the data into a category set and a numerical set, uses the stacking strategy, and combines it with catboost, decision tree, and extreme gradient boosting. During the feature engineering process, the ratio and classification features are added to the category feature set; the sales at t are used for training to obtain the prediction at (t + 1); and the features used in the prediction at time (t + 1) are generated by the prediction results at t. The update processes of the two sets are combined to form a joint feature update mechanism, and multiple features of k categories are jointly updated. Using this method, data of all categories of retail stores can be linked so that historical data of different categories of goods can provide support for the prediction of each category of goods and solve the problem of insufficient product data and features. The method is verified on the retail sales data obtained from the Kaggle platform, and the mean absolute error and weighted mean absolute percentage error are adopted to analyze the performance of the model. The results reveal that the forecasting method can provide a useful reference to decision-makers.

Multilayer Three‐Dimensional Woven Silver Nanowire Networks for Absorption‐Dominated Electromagnetic Interference Shielding

AbstractSecondary electromagnetic reflection is a great challenge in the field of high conductivity electromagnetic interference (EMI) material due to the poor electromagnetic absorption ability. In this work, a multilayer stack strategy compositing of the absorption‐dominated thin films and the reflection‐dominated thick film is proposed to realize both high shielding effectiveness (SE) and low reflection. The film comprises 3D woven silver nanowire (AgNWs) networks embedded in the polydimethylsiloxane (PDMS) substrate, and the shielding film thickness can be adjusted via controlling coating cycles. It can be found the three‐ply stack of two thin films and a thick film, (B:F)10(B:F)10(B:F)40, (B:F)5(B:F)10(B:F)40, and (B:F)5(B:F)5(B:F)40, can achieve EMI SEs up to 40.8, 40.7, and 36.3 dB, and the absorptions up to 0.59, 0.65, and 0.69, respectively. Results from verification investigation and theoretical simulation indicate that the multilayer can regulate the shielding mechanism of AgNWs‐based films from reflection‐dominated to absorption‐dominated, with no apparent deterioration of EMI SE. The synergistic effect of a multilayer stack is regarded to originate from a special “absorb‐absorb‐reflect‐absorb‐absorb” course. This finding suggests a promising approach to acquiring both high‐efficient and absorption‐dominated EMI shielding material.

Evaluation of Polymeric Particles for Modular Tissue Cultures in Developmental Engineering.

Developmental engineering (DE) aims to culture mammalian cells on corresponding modular scaffolds (scale: micron to millimeter), then assemble these into functional tissues imitating natural developmental biology processes. This research intended to investigate the influences of polymeric particles on modular tissue cultures. When poly(methyl methacrylate) (PMMA), poly(lactic acid) (PLA) and polystyrene (PS) particles (diameter: 5-100 µm) were fabricated and submerged in culture medium in tissue culture plastics (TCPs) for modular tissue cultures, the majority of adjacent PMMA, some PLA but no PS particles aggregated. Human dermal fibroblasts (HDFs) could be directly seeded onto large (diameter: 30-100 µm) PMMA particles, but not small (diameter: 5-20 µm) PMMA, nor all the PLA and PS particles. During tissue cultures, HDFs migrated from the TCPs surfaces onto all the particles, while the clustered PMMA or PLA particles were colonized by HDFs into modular tissues with varying sizes. Further comparisons revealed that HDFs utilized the same cell bridging and stacking strategies to colonize single or clustered polymeric particles, and the finely controlled open pores, corners and gaps on 3D-printed PLA discs. These observed cell-scaffold interactions, which were then used to evaluate the adaptation of microcarrier-based cell expansion technologies for modular tissue manufacturing in DE.

Application of a Fusion Model Based on Machine Learning in Visibility Prediction

To improve the accuracy of atmospheric visibility (V) prediction based on machine learning in different pollution scenarios, a new atmospheric visibility prediction method based on the stacking fusion model (VSFM) is established in this paper. The new method uses the stacking strategy to fuse two base learners—eXtreme gradient boosting (XGBoost) and light gradient boosting machine (LightGBM)—to optimize prediction accuracy. Furthermore, seasonal feature importance evaluations and feature selection were utilized to optimize prediction accuracy in different seasons with different pollution sources. The new VSFM was applied to 1-year environmental and meteorological data measured in Qingdao, China. Compared to other traditional non-stacking models, the new VSFM improved precision during different seasons, especially in extremely low-visibility scenarios (V< 2 km). The TS score of the VSFM was significantly better than that of other models. For extremely low-visibility scenarios, the VSFM had a threat score (TS) of 0.5, while the best performance of other models was less than 0.27. The new method is promising for atmospheric visibility prediction under complex urban pollution conditions. The research results can also improve our understanding of the factors that influence urban visibility.

Stacking strategy-assisted random forest algorithm and its application

Short-term power load forecasting provides important guidance for the improvement of power marketing and control levels of power enterprises. In this paper, a novel method, named RF-TStacking, is proposed to forecast the short-term load. This study starts from the influence factors of the power load, the random forest is applied to estimate the importance of the influence factors of short-term load. Based on Stacking strategy, the integration of LightGBM and random forest is realized to achieve short-term power load forecasting. To improve the generalization ability of the load model, random put back sampling is used to sample each primary learner, and the average value is taken as the result of each primary learner. The Bayesian optimization is used to adjust the super parameters of the model to improve the accuracy of the selection of influencing factors. The load data of a region in northwest China are used for the testing, and it is found that the model can provide stable prediction results.

Abstract P6-01-19: Multiparametric MRI-based Longitudinal-radiomics Analysis for Early Prediction of Treatment Response of Breast Cancers to Neoadjuvant Chemotherapy: A Multicenter Study

Abstract P6-01-19: Multiparametric MRI-based Longitudinal-radiomics Analysis for Early Prediction of Treatment Response of Breast Cancers to Neoadjuvant Chemotherapy: A Multicenter Study