Traditional Single Model Research Articles

A Vegetable-Price Forecasting Method Based on Mixture of Experts

The accurate forecasting of vegetable prices is crucial for policy formulation, market decisions, and agricultural market stability. Traditional time-series models often require manual parameter tuning and struggle to effectively handle the complex non-linear characteristics of vegetable price data, limiting their predictive accuracy. This study conducts a comprehensive analysis of the performance of traditional methods, deep learning approaches, and cutting-edge large language models in vegetable-price forecasting using multiple predictive performance metrics. Experimental results demonstrate that large language models generally outperform other methods, but do not have consistent performance for all kinds of vegetables across different time scales. As a result, we propose a novel vegetable-price forecasting method based on mixture of expert models (VPF-MoE), which combines the strengths of large language models and deep learning methods. Different from the traditional single model prediction method, VPF-MoE can dynamically adapt to the characteristics of different vegetable types, dynamically select the best prediction method, and significantly improve the accuracy and robustness of the prediction. In addition, we optimized the application of large language models in vegetable-price forecasting, offering a new technological pathway for vegetable-price prediction.

Multi-objective mixed-model assembly line balancing with hierarchical worker assignment: A case study of gear reducer manufacturing operations

Assembly lines, generally speaking, can reduce production costs, shorten cycle times, and achieve higher quality levels. Since the current market is characterized by increasing product variability, mixed-model assembly lines, in which similar product models can be assembled simultaneously, are more suitable to respond to varied market demands than traditional single-model assembly lines. In addition, in an assembly line, tasks often differ in processing requirements, and workers may have different qualification levels. This study, therefore, aims to construct models for the multi-objective mixed-model assembly line balancing problem with hierarchical worker assignment (MO-MALBP-HW). The goal is to generate a suitable plan for a mixed-model assembly line balancing problem considering the constraint of a hierarchical workforce, the cost of a hierarchical workforce, and production cycle time. When the problem is simple, it can be solved by a mixed integer programming (MIP) model. When the problem becomes complex, it can be solved by a multi-objective genetic algorithm (MOGA) and a non-dominated sorting genetic algorithm II (NSGA-II) to obtain a near-optimal solution. The implementation of this model can effectively manage the multi-objective mixed-model assembly line balancing plan, thereby improving plant efficiency and reducing cost.

Optimizing hypertension prediction using ensemble learning approaches

Hypertension (HTN) prediction is critical for effective preventive healthcare strategies. This study investigates how well ensemble learning techniques work to increase the accuracy of HTN prediction models. Utilizing a dataset of 612 participants from Ethiopia, which includes 27 features potentially associated with HTN risk, we aimed to enhance predictive performance over traditional single-model methods. A multi-faceted feature selection approach was employed, incorporating Boruta, Lasso Regression, Forward and Backward Selection, and Random Forest feature importance, and found 13 common features that were considered for prediction. Five machine learning (ML) models such as logistic regression (LR), artificial neural network (ANN), random forest (RF), extreme gradient boosting (XGB), light gradient boosting machine (LGBM), and a stacking ensemble model were trained using selected features to predict HTN. The models’ performance on the testing set was evaluated using accuracy, precision, recall, F1-score, and area under the curve (AUC). Additionally, SHapley Additive exPlanations (SHAP) was utilized to examine the impact of individual features on the models’ predictions and identify the most important risk factors for HTN. The stacking ensemble model emerged as the most effective approach for predicting HTN risk, achieving an accuracy of 96.32%, precision of 95.48%, recall of 97.51%, F1-score of 96.48%, and an AUC of 0.971. SHAP analysis of the stacking model identified weight, drinking habits, history of hypertension, salt intake, age, diabetes, BMI, and fat intake as the most significant and interpretable risk factors for HTN. Our results demonstrate significant advancements in predictive accuracy and robustness, highlighting the potential of ensemble learning as a pivotal tool in healthcare analytics. This research contributes to ongoing efforts to optimize HTN prediction models, ultimately supporting early intervention and personalized healthcare management.

Ensemble-based assisted history matching of DTS monitoring data in HT-ATES systems: application to a real-life case study

Underground heat storage is an important element in accelerating the energy transition. It can significantly contribute to CO2 emission reduction and cost savings since it is one of the cheapest forms of energy storage and it enables the seasonal storage of large energy surpluses from sustainable sources, e.g. wind, sun, geothermal. Numerical models are used for the prediction of thermal behavior important in establishing the high efficiency of the high temperature aquifer thermal energy storage (HT-ATES) systems. However, the lack of exact knowledge of the subsurface conditions introduces modeling uncertainty. It is therefore important to employ approaches that reduce subsurface uncertainty. History matching is a methodology where the numerical models are updated to match historical observations that will in turn not only increase understanding of the subsurface but also improve accuracy of the model predictability of future behavior. In this research, models of the first large-scale operational HT-ATES system in Middenmeer, the Netherlands, were used to evaluate the thermal evolution in the storage aquifer and the over- and underburden clay layers. The HT-ATES system, consisting of a hot and warm well, with a monitoring well inbetween, became operational in the summer of 2021. The extensive monitoring program implemented for the first few operational years provided an opportunity to study the performance of such a system from an environmental and operational point of view. A state-of-the-art assisted history matching approach was applied to the first storage cycle, using a coupling between history matching software and the thermal flow simulator. This approach was compared to a more traditional single-model manual history matching method. Rock properties of the aquifer and over- and underburden layers were updated in the randomly generated prior ensemble of models to fit the simulated temperature evolution measured down the monitoring well with the distributed temperature sensing (DTS) data. The observations gathered during the second year of operations were used to validate the accuracy of the prediction capabilities of the updated models. The obtained results indicate the value of history matching to improve understanding of the subsurface conditions for HT-ATES systems and obtain models with better predictability of the future behavior of heat in the storage reservoir and overburden formations. Such improved models are instrumental in providing engineers with a better quantitative grip on the environmentally responsible storage potential and heat deliverability of the target storage site, which is important to achieve cost-effective site-specific design (e.g. number of wells, well placement) and performing operational strategies (e.g. injection/production rates and temperatures) for new HT-ATES systems. Moreover, the benefits of the assisted history matching approach over manual method are highlighted and both approaches are validated where the assisted history matching method produced more accurate predictions than the manual approach.

A Novel Workflow for Mapping Forest Canopy Height by Synergizing ICESat-2 and Multi-Sensor Data

Precise information on forest canopy height (FCH) is critical for forest carbon stocks estimation and management, but mapping continuous FCH with satellite data at regional scale is still a challenge. By fusing ICESat-2, Sentinel-1/2 images and ancillary data, this study aimed to develop a workflow to obtain an FCH map using a machine learning algorithm over large areas. The vegetation-type map was initially produced by a phenology-based spectral feature selection method. A forest characteristic-based model was then proposed to map spatially continuous FCH after a multivariate quality control. Our results show that the overall accuracy (OA) and average F1 Score (F1) for eight main vegetation types were more than 90% and 89%, respectively, and the vegetation-type map agreed well with the census areas. The forest characteristic-based model demonstrated a greater potential in FCH prediction, with an R-value 60.47% greater than the traditional single model, suggesting that the addition of the multivariate quality control and forest structure characteristics could positively contribute to the prediction of FCH. We generated a 30 m continuous FCH map by the forest characteristic-based model and evaluated the product with about 35 km2 of airborne laser scanning (ALS) validation data (R = 0.73, RMSE = 2.99 m), which were 45.34% more precise than the China FCH, 2019. These findings demonstrate the potential of our proposed workflow for monitoring regional continuous FCH, and will greatly benefit accurate forest resources assessment.

Multimodel-based quantitative source apportionment and risk assessment of soil heavy metals: A reliable method to achieve regional pollution traceability and management

To strengthen the control of pollution sources and promote soil pollution management of agricultural land, this study constructed a comprehensive source apportionment framework, which significantly improved the reliability of potential source analysis compared with the traditional single model. The spatial distribution pattern of agricultural soil heavy metals (SHMs) content in Lintong, a typical river valley city in China, was determined and the degree of contamination was evaluated. A scientific source apportionment methodological framework was constructed through correlation analysis methods together with multiple source apportionment receptor models. Finally, the Monte Carlo simulation method was used to derive the results of the human health risk assessment (HHRA). The results revealed the following: (1) Agricultural soils were moderately and mildly polluted, accounting for 28.8 % and 71.2 % of the total number of sampling points, respectively. (2) The overall correlation of heavy metals (HMs) was strong according to the coupling analysis of the SHMs, in which a strong correlation (0.8–1) was reached among Cu, Ni, Pb, Cr and Zn, indicating that these HMs were most likely homologous or composite. (3) Multimodel analysis of the SHMs sources revealed that the first and second principal components were agricultural (41.36 %) and industrial (19.69 %) sources, respectively, and the remaining principal components were road traffic, natural factors, and atmospheric deposition or surface runoff, respectively. (4) The average comprehensive noncarcinogenic health risk indices for adults and children were 4.2259E-02 and 1.4194E-01, respectively, which were within the slight risk range, indicating that the risk caused by SHMs to the human body can be almost negligible. This study adopted a mixed method to reveal the risk of SHMs pollution and its sources, which provides some reference and technical support for traceability analysis, zoning control, and health risk studies of regional pollutants and is helpful for formulating scientific management measures and targeting control policies.

Intelligent hardware design for wire status monitoring integrating Beidou GNSS and inertial measurement technology in machine learning environment

ABSTRACT Conventional wire condition monitoring methods often lack the precision and reliability required for real-time assessment in complex environments, leading to undetected failures and compromised power transmission safety. To address these limitations, the article presents an XGBoost (Extreme Gradient Boosting) – LSTM (Long and Short-Term Memory) hybrid model for wire state monitoring using Beidou GNSS (Global Navigation Satellite System) and IMU (Inertial Measurement Unit) sensors. The system tracks wire vibration, tilt, rotation, bending, and stretching in real time. GNSS provides positional data, and IMU records acceleration and angular velocity. The Kalman filter cleans data, and XGBoost and LSTM are used for feature selection and temporal modelling. The combined model achieves 94.1% accuracy overall, with XGBoost at 92.4% and LSTM at 90.9%. Classification accuracy is 95.5% for vibration, 94.7% for bending, and 94.3% for stretching. The experiment also showed that the model has stable performance in different temperature environments, with accuracy fluctuating between 94.0% and 94.2% within the temperature range of −5℃ to 50℃. In general, the XGBoost LSTM model has achieved high-precision and high stability monitoring in wire condition monitoring, which is significantly better than the traditional single model. This study improves the accuracy of wire condition monitoring and has significant practical application value and strong stability.

New energy vehicle battery state of charge prediction based on XGBoost algorithm and RF fusion

As the most important component of new energy electric vehicles, lithium-ion batteries may suffer irreversible damage to the battery due to an abnormal state of charge. Nevertheless, the extant research on charge prediction predominantly employs a single model or an enhanced single model. However, these approaches do not fully account for the intricacies and variability of the actual driving road conditions of the vehicle. Furthermore, the prediction accuracy of the charge state in the latter phase of discharge remains suboptimal. To further improve the accuracy of predicting the state of charge, the study utilizes actual operating data of new energy vehicles and combines two proposed algorithms to build a first layer learner of a fusion prediction model. The second layer learner integrates various prediction results. The fusion model can enhance its adaptability to complex data structures by combining the gradient boosting ability of XGBoost algorithm and the diversity of Random Forest when dealing with nonlinear problems. This fusion method modifies the input features of the second layer of the fusion model, enhances the complexity of the second layer learner, effectively circumvents overfitting, and exhibits reduced error rates relative to traditional single-chip prediction models. As a result, the performance of the prediction model is significantly enhanced. The tests showed that when using the fusion model for state of charge prediction, the prediction accuracy could reach 97.6%, and the prediction accuracy was higher than the other four comparison models. When the car was driving in a 25 ℃ highway environment, the root mean square error of the fusion model was 1.3%, and the average absolute error was 1.5%. In urban road environments, the root mean square error of the fusion model was 1.5%, and the average absolute error was 1%. The experiment demonstrates that the proposed fusion prediction model can accurately predict the charging status, thereby enabling the battery to be fully utilized while simultaneously reducing energy consumption. In comparison to the traditional single model or enhanced single model, the proposed fusion model has demonstrated a notable enhancement in both prediction accuracy and computational efficiency.

Estimation of earthquake-induced direct economic losses of portfolio buildings based on seismic fragility surface

Regional seismic loss assessment is crucial for developing earthquake emergency plans, which can significantly reduce casualties and socioeconomic losses in urban communities. Due to the complex types of building structures in a region, it is not possible to accurately measure the damage state of building structures using a single scalar ground motion intensity measure (IM). To more accurately reflect the responses of structures under earthquake excitations, the seismic fragility surface models for 4 typical RC frame structures are developed based on the vector ground motion IMs. The results indicate that the fragility surface model offers a superior level of precision in capturing the inherent uncertainty of ground motion and subsequently determines a more precise probability of structural failure, outperforming traditional single scalar ground motion IM fragility curve models. This enhanced accuracy holds significant implications for regional loss assessment. To fully consider the spatial cross-correlation of vector ground motion IMs, based on the established ground motion database of 8778 records, principal component analysis (PCA) and geostatistical analysis methods are used to construct a spatial cross-correlation influence field that simultaneously considers multiple ground motion IMs to simulate more realistic earthquake scenarios. Considering the impact of uncertainty in loss ratio on earthquake-induced direct economic loss estimation results, the maximum entropy technique is used in conjunction with an improved Latin Hypercube Sampling (LHS) to achieve spatially uniform sampling of loss ratio (LR). This study proposes an assessment framework for regional seismic direct economic loss for portfolio buildings based on the seismic fragility surface model considering spatial cross-correlation and uncertainty of loss ratio. The proposed framework is verified by numerical examples, and it is proven that the regional portfolio buildings considering multiple ground motion IMs is more reliable and robust than those merely considering a single ground motion IM, and when spatial cross-correlation is not considered, the probability of serious economic losses will be underestimated.

Multi-Timescale Energy Consumption Management in Smart Buildings Using Hybrid Deep Artificial Neural Networks

Demand side management is a critical issue in the energy sector. Recent events such as the global energy crisis, costs, the necessity to reduce greenhouse emissions, and extreme weather conditions have increased the need for energy efficiency. Thus, accurately predicting energy consumption is one of the key steps in addressing inefficiency in energy consumption and its optimization. In this regard, accurate predictions on a daily, hourly, and minute-by-minute basis would not only minimize wastage but would also help to save costs. In this article, we propose intelligent models using ensembles of convolutional neural network (CNN), long-short-term memory (LSTM), bi-directional LSTM and gated recurrent units (GRUs) neural network models for daily, hourly, and minute-by-minute predictions of energy consumptions in smart buildings. The proposed models outperform state-of-the-art deep neural network models for predicting minute-by-minute energy consumption, with a mean square error of 0.109. The evaluated hybrid models also capture more latent trends in the data than traditional single models. The results highlight the potential of using hybrid deep learning models for improved energy efficiency management in smart buildings.

A River Water Quality Prediction Method Based on Dual Signal Decomposition and Deep Learning

Traditional single prediction models struggle to address the complexity and nonlinear changes in water quality forecasting. To address this challenge, this study proposed a coupled prediction model (RF-TVSV-SCL). The model includes Random Forest (RF) feature selection, dual signal decomposition (Time-Varying Filtered Empirical Mode Decomposition, TVF-EMD, and Sparrow Search Algorithm-Optimized Variational Mode Decomposition, SSA-VMD), and a deep learning predictive model (Sparrow Search Algorithm-Convolutional Neural Network-Long Short-Term Memory, SSA-CNN-LSTM). Firstly, the RF method was used for feature selection to extract important features relevant to water quality prediction. Then, TVF-EMD was employed for preliminary decomposition of the water quality data, followed by a secondary decomposition of complex Intrinsic Mode Function (IMF) components using SSA-VMD. Finally, the SSA-CNN-LSTM model was utilized to predict the processed data. This model was evaluated for predicting total phosphorus (TP), total nitrogen (TN), ammonia nitrogen (NH3-N), dissolved oxygen (DO), permanganate index (CODMn), conductivity (EC), and turbidity (TB), across 1, 3, 5, and 7-d forecast periods. The model performed exceptionally well in short-term predictions, particularly within the 1–3 d range. For 1-, 3-, 5-, and 7-d forecasts, R2 ranged from 0.93–0.96, 0.79–0.87, 0.63–0.72, and 0.56–0.64, respectively, significantly outperforming other comparison models. The RF-TVSV-SCL model demonstrates excellent predictive capability and generalization ability, providing robust technical support for water quality forecasting and pollution prevention.

Hybrid wind speed optimization forecasting system based on linear and nonlinear deep neural network structure and data preprocessing fusion

Wind speed time series forecasting has been widely used in wind power generation. However, the nonlinear and non-stationary characteristics of wind speed make accurate wind speed forecasting a difficult task. In recent years, the rapid development of artificial intelligence and machine learning technology provides a new solution to the problem of wind speed forecasting. Combining the advance of artificial intelligence and data analysis strategy, this paper proposes a wind speed forecasting system based on multi-model fusion and integrated learning. Considering the differences in data observation and training principles of various algorithms and exploiting the advantages of each model, a wind speed forecasting system with multiple machine learning algorithms embedded in integrated learning is constructed, which includes three modules: data preprocessing, optimization and forecasting. The data preprocessing module can conduct quantitative analysis through input data decomposition and feature extraction, and the combination of multi-objective intelligent optimization algorithm and combined forecasting method can effectively forecast the wind speed time series. The validity of the algorithm is verified using the data of Shandong wind farm in China. The forecasting results show that compared with the traditional single model forecasting, the proposed integrated wind speed forecasting system based on the fusion of multiple models has higher forecasting accuracy.

AFC-Unet: Attention-fused full-scale CNN-transformer unet for medical image segmentation

In the field of medical image segmentation, although U-Net has achieved significant achievements, it still exposes some inherent disadvantages when dealing with complex anatomical structures and small targets, such as inaccurate target localization, blurry edges, and insufficient integration of contextual information. To address these challenges, this study proposes the Attention-Fused Full-Scale CNN-Transformer Unet (AFC-Unet), aiming to effectively overcome the limitations of traditional U-Net through multi-scale feature fusion, attention mechanisms, and CNN-Transformer hybrid modules. Specifically, we adopt an encoder–decoder architecture, incorporating full-scale feature block fusion and pyramid sampling modules to enhance the model’s ability to recognize fine to overall structural features by integrating cross-level multi-scale features. We propose the Multi-feature Fusion Attention Gates (MFAG) module, which focuses on and highlights discriminative information of potential lesions and key anatomical boundaries, effectively suppressing irrelevant background interference. We design a module Convolutional Hybrid Attention Transformer (CHAT) that integrates CNN and Transformer to address the shortcomings of traditional single models in handling long-range dependencies and global context understanding. Experimental results on three datasets of different scales demonstrate that the model’s segmentation performance for medical images surpasses state-of-the-art models, showcasing high generalization ability.

A Novel Fuzzy Logic Switched MPC for Efficient Path Tracking of Articulated Steering Vehicles

This paper introduces a novel fuzzy logic switched model predictive control (MPC) algorithm for articulated steering vehicles, addressing significant path tracking challenges due to varying road conditions and vehicle speeds. Traditional single-model and parameter-based controllers struggle with tracking errors and computational inefficiencies under diverse operational conditions. Therefore, a kinematics-based MPC algorithm is first developed, showing strong real-time performance but encountering accuracy issues on low-adhesion surfaces and at high speeds. Then, a 4-DOF dynamics-based MPC algorithm is designed to enhance tracking accuracy and control stability. The proposed solution is a switched MPC strategy, integrating a fuzzy control system that dynamically switches between kinematics-based and dynamics-based MPC algorithms based on error, solution time, and heading angle indicators. Subsequently, simulation tests are conducted using SIMULINK and ADAMS to verify the performance of the proposed algorithm. The results confirm that this fuzzy-based MPC algorithm can effectively mitigate the drawbacks of single-model approaches, ensuring precise, stable, and efficient path tracking across diverse adhesion road conditions.

Construction of Ensemble Learning Model for Home Appliance Demand Forecasting

Given the increasing competition among household appliance enterprises, accurately predicting household appliance demand is crucial for enterprise supply chain management and marketing. This paper proposes a combined model integrating deep learning and ensemble learning—LSTM-RF-XGBoost—to assist enterprises in identifying customer demand, thereby addressing the complexity and uncertainty of the household appliance market demand. In this study, Long Short-Term Memory Network (LSTM), Random Forest (RF), and Extreme Gradient Boosting (XGBoost) models are established separately. Then, the three individual algorithms are used as the base models in the first layer, with the multiple linear regression (MLR) algorithm serving as the meta-model in the second layer, merging the demand prediction model based on the hybrid model into the overall demand prediction model. This study demonstrates that the accuracy and stability of demand prediction using the LSTM–RF–XGBoost model significantly outperform traditional single models, highlighting the significant advantages of using a combined model. This research offers practical and innovative solutions for enterprises seeking rational resource allocation through demand prediction.

WVETT-Net: A Novel Hybrid Prediction Model for Wireless Network Traffic Based on Variational Mode Decomposition

Precise prediction of wireless communication network traffic is indispensable in the operational deployment of base station resources and improvement of the user experience. Cellular wireless network traffic has both spatial and temporal characteristics. The existing modeling algorithms have achieved good results in extracting the spatial features, but there are still deficiencies in the extraction models for the time dependencies. To resolve these problems, this paper proposes a novel hybrid neural network prediction model, called WVETT-Net. Firstly, variational mode decomposition (VMD) is used to preprocess network traffic, and the whale optimization algorithm (WOA) is used to select the optimal parameters for VMD. Secondly, the local and global features are extracted from each subsequence by a temporal convolutional network (TCN) and an improved Transformer network with a multi-head ProbSparse self-attention mechanism (Pe-Transformer), respectively. Finally, the extracted feature representation is enhanced by using an efficient channel attention (ECA) mechanism to achieve accurate wireless network traffic predictions. Experimental results on two wireless network traffic datasets show that the proposed model (WVETT-Net) outperforms the traditional single or combined models in wireless network traffic prediction.

A stacking ensemble model for predicting soil organic carbon content based on visible and near-infrared spectroscopy

The content of soil organic carbon (SOC) plays an important role in maintaining ecosystem functions, protecting soil biodiversity, and understanding carbon cycling processes. The combination of visible and near-infrared spectroscopy (VIS–NIRS) and machine learning can achieve rapid prediction of SOC content. However, it is still relatively unknown how to integrate the characteristics of various machine learning models to improve the performance of SOC prediction models. In this study, a new model for predicting SOC content based on stacking ensemble learning was proposed by using VIS–NIRS. The prediction performances of six different models including Support Vector Regression (SVR), Extreme Gradient Boosting (XGBoost), Random Forest (RF), Light Gradient Boosting Machine (LightGBM), PartialLeast-square (PLS) and Extreme Learning Machine (ELM) on SOC content under different spectral preprocessing methods were compared. The results indicated that SVR, XGBoost, and LightGBM models provide better prediction performance after first-order derivative preprocessing. After comparing the performance of various combinations of base models applied to the first layer of a stacking ensemble model, the results showed that both the combination of XGBoost, LightGBM, and SVR models and the combination of SVR, ELM, and LightGBM models achieve the best performance. The coefficient of determination (R2) of the stacking ensemble model on the test set reaches 0.84, which improves the accuracy of the model compared with the traditional single model. The stability of the stacking ensemble model was verified by applying it to datasets of different sizes, which can replace traditional machine learning models in predicting SOC content.

A Multi-model Fusion Strategy for Android Malware Detection Based on Machine Learning Algorithms

In the digital age, the widespread use of Android devices has led to a surge in security threats, especially malware. Android, as the most popular mobile operating system, is a primary target for malicious actors. Conventional antivirus solutions often fall short in identifying new, modified, or zero-day attacks. To address this, researchers have explored various approaches for Android malware detection, including static and dynamic analysis, as well as machine learning (ML) techniques. However, traditional single-model ML approaches have limitations in generalizing across diverse malware behaviors. To overcome this, a multi-model fusion approach is proposed in this paper. The approach integrates multiple machine learning models, including logistic regression, decision tree, and K-nearest neighbors, to improve detection accuracy. Experimental results demonstrate that the fusion method outperforms individual models, offering a more balanced and robust approach to Android malware detection. This methodology showcases the potential of ensemble techniques in enhancing prediction accuracy, providing valuable insights for future research in cybersecurity.

A prediction model for water absorption in sublayers based on stacking ensemble learning method

The water absorption profile can reflect the difference in the waterflooding status of each sublayer in the injection well, and predicting the water absorption profile has important guiding significance for formulating and adjusting the waterflood development plan. Currently, numerical simulation and radioisotope logging techniques are commonly used to obtain the water absorption in sublayers. However, the complex and increasingly heterogeneous subsurface formations in China's oilfields have made it impossible for the numerical simulation to reflect the actual reservoir water injection status by splitting the water absorption into sublayers. Although the radioisotope method is more accurate, it is more expensive and has less actual test data in oilfields. In addition, the water absorption prediction model built by traditional machine learning only applies to injection wells with a small amount of water absorption profile data, and its prediction accuracy needs further improvement. In this paper, a new machine learning prediction method is proposed to build a water absorption profile prediction model by stacking ensemble learning algorithm, which achieves the prediction of water absorption in sublayers for injection wells without historical data of water absorption profiles during the development period. In our model, based on the oilfield's static geological data and dynamic production data, the input feature parameters are comprehensively selected by a hybrid feature selection method to analyze the importance of each feature. The robust base-learners are extracted from the sub-model pool by comparing and analyzing the comprehensive prediction performance of each individual model, including ANN, KNN, SVR, RF, XGBoost, LightGBM, etc. The stacking ensemble model was trained using K-fold cross-validation on a randomly divided dataset, which improves the predictive performance of an individual model by combining the advantages of multiple base-learners so that it can accurately reflect the dynamic changes of the water absorption profile. Finally, a case study from the MG study area of the DG oilfield was conducted to verify the model's predictive effect. This case shows that the stacking model proposed in this paper has better performance in predicting the water absorption in sublayers compared with the traditional single model and other ensemble models, and the model has a strong generalization ability, which proves that the method can be used in studying the prediction of the water absorption in sublayers of injection wells.

Application of a Combination of Two Teaching Methods in Pediatric Teaching

Pediatrics is one of the clinical professional courses that is highly practical. The traditional single teaching method is inconducive to stimulating students’ interest in learning, resulting in a poor classroom teaching effect. To change the traditional single teaching model and promote and improve the development of pediatric teaching, according to the level of pediatric teaching, colleges can consider combining two teaching methods to carry out pediatric teaching. The main research in this paper is the combination of PBL (problem-based learning) teaching and CBS (case-based study) teaching. An overview of the two teaching methods is first given, and then the application significance of the combination of the two teaching methods is analyzed. Lastly, it explores the specific implementation of hybrid teaching methods in pediatric teaching.