Single Model Research Articles

Multiobjective Combinatorial Optimization Using a Single Deep Reinforcement Learning Model.

This article proposes utilizing a single deep reinforcement learning model to solve combinatorial multiobjective optimization problems. We use the well-known multiobjective traveling salesman problem (MOTSP) as an example. Our proposed method employs an encoder-decoder framework to learn the mapping from the MOTSP instance to its Pareto-optimal set. Specifically, it leverages a novel routing encoder to extract information for both the entire multiobjective aspect and every individual objective from the MOTSP instance. The global embeddings and each objective's embeddings are adaptively aggregated via a routing network to form the subproblems' embedding that can well represent the MOTSP features. Using a modified context embedding, the subproblems' embeddings are fed into a decoder to produce a set of approximate Pareto-optimal solutions in parallel. Additionally, we develop a Top-k baseline to enable more efficient data utilization and lightweight training for our proposed method. We compare our method with heuristic-based and learning-based ones on various types of MOTSP instances, and the experimental results show that our method can solve MOTSP instances in real-time and outperform the other algorithms, especially on large-scale problem instances.

Bare-Metal Stent Tracking with Magnetic Particle Imaging.

Magnetic particle imaging (MPI) is an emerging medical imaging modality that is on the verge of clinical use. In recent years, cardiovascular applications have shown huge potential like, e.g., intraprocedural imaging guidance of stent placement through MPI. Due to the lack of signal generation, nano-modifications have been necessary to visualize commercial medical instruments until now. In this work, it is investigated if commercial interventional devices can be tracked with MPI without any nano-modification. Potential MPI signal generation of nine endovascular metal stents was tested in a commercial MPI scanner. Two of the stents revealed sufficient MPI signal. Because one of the two stents showed relevant heating, the imaging experiments were carried out with a single stent model (Boston Scientific/Wallstent-Uni Endoprothesis, diameter: 16 mm, length: 60 mm). The nitinol stent and its delivery system were investigated in seven different scenarios. Therefore, the samples were placed at 49 defined spatial positions by a robot in a meandering pattern during MPI scans. Image reconstruction was performed, and the mean absolute errors (MAE) between the signals' centers of mass (COM) and ground truth positions were calculated. The stent material was investigated by magnetic particle spectroscopy (MPS) and vibrating sample magnetometry (VSM). To detect metallic components within the delivery system, nondestructive testing via computed tomography was performed. The tracking of the stent and its delivery system was possible without any nano-modification. The MAE of the COM were 1.49 mm for the stent mounted on the delivery system, 3.70 mm for the expanded stent and 1.46 mm for the delivery system without the stent. The results of the MPS and VSM measurements indicate that besides material properties eddy currents seem to be responsible for signal generation. It is possible to image medical instruments with dedicated designs without modifications by means of MPI. This enables a variety of applications without compromising the mechanical and biocompatible properties of the instruments.

Combinatorial reasoning-based abnormal sensor recognition method for subsea production control system

The subsea production system is a vital equipment for offshore oil and gas production. The control system is one of the most important parts of it. Collecting and processing the signals of subsea sensors is the only way to judge whether the subsea production control system is normal. However, subsea sensors degrade rapidly due to harsh working environments and long service time. This leads to frequent false alarm incidents. A combinatorial reasoning-based abnormal sensor recognition method for subsea production control system is proposed. A combinatorial algorithm is proposed to group sensors. The long short-term memory network (LSTM) is used to establish a single inference model. A counting-based judging method is proposed to identify abnormal sensors. Field data from an offshore platform in the South China Sea is used to demonstrate the effect of the proposed method. The results show that the proposed method can identify the abnormal sensors effectively.

Two-Phase Adaptive Kriging Model Based Importance Sampling Method for Estimating Time-Dependent Failure Probability

To improve the computational efficiency of the time-dependent failure probability (TDFP) to system with dynamic loads, materials degeneration and other time-varying factors, a two-phase adaptive Kriging model-based importance sampling (IS) method is proposed. The main contribution of this article is that a single Kriging model is established and adaptively refined to gradually approach the optimal IS probability density function (PDF) and estimate TDFP in two sequential phases. In the first phase, a generally approximate IS PDF is constructed by Kriging model based on its prediction characteristic and subsequently updated until convergence. Then, a simple rejection sampling technique is used to robustly generate IS samples for the estimation of TDFP. In the second phase, current Kriging model is subsequently refined for accurately identifying the classification of the IS samples generated in the last phase. Finally, the TDFP can be estimated using the products obtained in the above two phases. As the IS PDF established by Kriging model is used for directly approximating the optimal one, the proposed method has the capacity to deal with the complicated problem containing multiple failure regions. Additionally, since IS procedure can dramatically improve the sampling efficiency, the proposed method is an effective technique to the time-consuming TDFP analysis problems.

Rapid Review of Therapy Protocols for Public Health Emergencies.

How research during a public health emergency is conducted is recognized as essential to the public health response to that emergency. Such research needs to undergo substantive and meaningful ethical review in a timely manner. Rapid ethical review may be accomplished through a number of mechanisms, including use of local rapid-response institutional review boards (IRBs). We describe use of such a model in the setting of the 2014 Ebola virus disease epidemic and the Rapid-Response IRB's subsequent transition to a multisite single IRB model during the current Covid-19 pandemic. The rapid-response review model is characterized by a small IRB with extensive use of alternate members with specific expertise and by close collaboration with the investigator in an iterative process.

Improved reference condition independent method for output performance estimation of PV modules under varying operating conditions

Traditional methods for estimating output property of the photovoltaic (PV) modules are strongly influenced by the selection of reference condition and transforming equations, which determine the calculated physical parameters under real operating conditions. The differences in the carrier transport properties of PV cells under varying operating conditions, such as the number and velocity of minority carriers at the junction edge and their recombination speed, lead to large deviations in the estimation of the output characteristics, especially under low irradiance conditions. To enhance the accuracy of performance estimation, we propose an improved method that is independent of reference condition. This method eliminates the impact of reference conditions and improves the transformation equations under all irradiance levels. Transformation equations of single diode model are established in different irradiance intervals based on the dependence of physical parameter on irradiance and temperature. Especially in the low irradiance range, all effects of irradiance and temperature are considered for each physical parameter in improved transformation equations. To optimize the unknown parameters in the transformation equations, the artificial hummingbird algorithm is used to fit experimental I–V data. The experimental results of six different types PV modules under a wide range of operating conditions are used to verify the effectiveness of the proposed method. The proposed method offers immediate benefits, including independence from reference condition and a more precise relationship between physical parameters and environmental factors in the estimation of PV output properties. Comparing the results to the traditional method by Laudani, the proposed method demonstrates superior capability in estimating I–V characteristics and accurately identifies the maximum power point under various operating conditions, which is of significant value for engineering applications.

An adaptive weight search method based on the Grey wolf optimizer algorithm for skin lesion ensemble classification

AbstractSkin cancer is a common type of malignant tumor that poses a serious threat to patients' lives and health, especially melanoma. It may spread to other body parts, resulting in serious complications and death. In the medical field, accurate identification of skin lesion images is crucial for diagnosing different diseases. However, due to the similarity between different skin lesions, it brings some challenges to medical diagnosis. In this paper, a novel Ensemble Learning Model (EL‐DLOA) based on deep learning and optimization algorithms is proposed, which uses four different deep neural network architectures to generate confidence levels for classes, and optimization algorithms are used to integrate these confidence levels to make the final predictions. To ensure the model's accuracy and reliability, it is first trained using three different learning rates to find the best classification performance of the model. Then, a new search method based on the grey wolf optimization algorithm is proposed to enhance the grey wolf search efficiency. The method improves the search mechanism by changing the grey wolf's individual position through random perturbation or adaptive mutation, which solves the problem that the grey wolf algorithm is easy to fall into local optimum. Finally, four different ensemble strategies are used to reduce individual model bias in the classification process. The proposed model is trained and evaluated using the publicly available dataset HAM10000. The experimental results show that the improved grey wolf optimization algorithm effectively avoids the premature convergence problem and improves the search combination efficiency. Furthermore, in the ensemble methods, the adaptive weight average ensemble strategy effectively improves the classification performance, yielding accuracy, precision, recall, and F1 scores of 0.888, 0.837, 0.897, and 0.862, respectively. These metrics show varying degrees of improvement over the best performing single model. In general, the results indicate that the proposed method achieves high accuracy and practicality in skin lesion classification. Our model shows excellent performance in comparison with other existing models, which makes it significant for research and application in dermatology diagnosis.

Intra- and peritumoral radiomics features based on multicenter automatic breast volume scanner for noninvasive and preoperative prediction of HER2 status in breast cancer: a model ensemble research

The aim to investigate the predictive efficacy of automatic breast volume scanner (ABVS), clinical and serological features alone or in combination at model level for predicting HER2 status. The model weighted combination method was developed to identify HER2 status compared with single data source model method and feature combination method. 271 patients with invasive breast cancer were included in the retrospective study, of which 174 patients in our center were randomized into the training and validation sets, and 97 patients in the external center were as the test set. Radiomics features extracted from the ABVS-based tumor, peritumoral 3 mm region, and peritumoral 5 mm region and clinical features were used to construct the four types of the optimal single data source models, Tumor, R3mm, R5mm, and Clinical model, respectively. Then, the model weighted combination and feature combination methods were performed to optimize the combination models. The proposed weighted combination models in predicting HER2 status achieved better performance both in validation set and test set. For the validation set, the single data source model, the feature combination model, and the weighted combination model achieved the highest area under the curve (AUC) of 0.803 (95% confidence interval [CI] 0.660–947), 0.739 (CI 0.556,0.921), and 0.826 (95% CI 0.689,0.962), respectively; with the sensitivity and specificity were 100%, 62.5%; 81.8%, 66.7%; 90.9%,75.0%; respectively. For the test set, the single data source model, the feature combination model, and the weighted combination model attained the best AUC of 0.695 (95% CI 0.583, 0.807), 0.668 (95% CI 0.555,0.782), and 0.700 (95% CI 0.590,0.811), respectively; with the sensitivity and specificity were 86.1%, 41.9%; 61.1%, 71.0%; 86.1%, 41.9%; respectively. The model weighted combination was a better method to construct a combination model. The optimized weighted combination models composed of ABVS-based intratumoral and peritumoral radiomics features and clinical features may be potential biomarkers for the noninvasive and preoperative prediction of HER2 status in breast cancer.

Research on frequency-variation square ratio in damage identification technique for single column platforms

This paper attempts to utilize lower order harmonics and modes of vibration consisting of “square ratios of frequency variations” as structural damage susceptibility characteristics. This is followed by the development of the frequency-based method. Firstly, this paper derives the theoretical formulation of the square ratio of frequency variation method. Secondly, a finite element model of a single column platform is established. Then the damage is simulated using the reduced modulus of elasticity to localize the damage. The result is shown to be feasible. Thirdly, a model of the offshore single column platform is constructed, and the damage is simulated using cracks on the column. The “frequency variation square ratio” is valid, so the method can be used in reality. This method is applied to the single column platform model, suitable for a beam structure. It is very effective for the single column damage identification at different damage levels, and can also initially locate the damage. This method has high practical application value and reference value for future studies.

Experimental Study of Wake Evolution under Vertical Staggered Arrangement of Wind Turbines of Different Sizes

During the expansion of a wind farm, the strategic placement of wind turbines can significantly improve wind energy utilization. This study investigates the evolution of wake turbulence in a wind farm after introducing smaller wind turbines within the gaps between larger ones, focusing on aspects such as wind speed, turbulence intensity, and turbulence integral length scale. The flow field conditions are described using parameters like turbulence critical length and power spectral density, as determined through wind tunnel experiments. In these experiments, a single large wind turbine model and nine smaller wind turbine models were used to create a small wind farm unit, and pressure distribution behind the wind turbines was measured under various operating conditions. The results indicate that downstream wind speed deficits intensify as the number of small wind turbines in operation increases. The impact of these smaller turbines varies with height, with a relatively minor effect on the upper blade tip and increasingly adverse effects as you move from the upper blade tip to the lower blade tip. Through an analysis of power spectral density, the contribution of vortex motion to wake turbulence kinetic energy is further quantified. In the far wake region, the number of small wind turbines has a relatively small impact on wind speed fluctuations.

Determination of Optimal Stock Portfolio Return by Single Index Model (Case Study on Banking Sector Stocks in Indonesia)

The optimal portfolio is a portfolio chosen by investors from the many options available in the collection of efficient portfolios. To get the optimal proportion, which is the maximum return and minimum risk, it is necessary to analyze the stocks to be selected in the investment model. The research objective is to determine the optimal return, risk, and proportion for each banking stock portfolio in Indonesia in the period February - July 2023. The method used is the Single Index Model. The process of determining the optimal proportion of stocks with the Single Index Model requires stock and market return data as the main basis for applying this method. This study involves the formation of an optimal portfolio of daily closing prices of 46 banking stocks. As a result of this research, there are 5 optimal stocks that meet the criteria for optimal portfolio formation with each fund proportion of 21.43% (BNII), 13.52% (BDMN), 35.02% (BBRI), 23.69% (BTPN), and 6.34% (BBCA). Expected return from optimal stocks is 0.152% and the risk that will be borne by investors is 0.0011% per day.

Robust planning upon unit commitment

Introducing operational flexibility in power system planning has motivated the development of planning models to be built upon and validated by unit commitment. While handling uncertainty, these models form a holistic robust optimization that makes three groups of decisions in a single model, namely investment, commitment, and dispatch. A dilemma arises when aligning commitment with other decisions, either on the same layer with investment or with dispatch—the former makes the model easy to solve but violates the decision-making process in reality, while the latter creates a triple-layer optimization with a mixed-integer inner layer that is harder to solve. A tractable routine is proposed for the latter with a nested decomposition algorithm to contribute a technique that splits the investment and commitment decisions for power system planning under uncertainty, which is well suited to the real-world infrastructure investment decision process.

Significant duration prediction of seismic ground motions using machine learning algorithms.

This study aims to predict the significant duration (D5-75, D5-95) of seismic motion by employing machine learning algorithms. Based on three parameters (moment magnitude, fault distance, and average shear wave velocity), two additional parameters(fault top depth and epicenter mechanism parameters) were introduced in this study. The XGBoost algorithm is utilized for characteristic parameter optimization analysis to obtain the optimal combination of four parameters. We compare the prediction results of four machine learning algorithms (random forest, XGBoost, BP neural network, and SVM) and develop a new method of significant duration prediction by constructing two fusion models (stacking and weighted averaging). The fusion model demonstrates an improvement in prediction accuracy and generalization ability of the significant duration when compared to single algorithm models based on evaluation indicators and residual values. The accuracy and rationality of the fusion model are validated through comparison with existing research.

Hybrid wavelet transform – MLR and ANN models for river flow prediction: Case study of Brahmaputra river (Pancharatna station)

In this research, discrete wavelet transform (DWT) is combined with MLR and ANN to develop WMLR and WANN hybrid models, respectively, for the Brahmaputra river (Pancharatna station) flow forecasting. Daily flow data for the period of 10 year were decomposed (up to fifth level) into detailed and approximation coefficients (using Daubechies wavelets db1, db2, db3, db8 and db10) which were fed as input to MLR and ANN to get the predicted discharge values two days, four days, seven days and 14 days ahead. For all lead times, the WMLR-db10 model was found to be superior as compared to WANN-db1, WANN-db2, WANN-db3, WANN-db8, WMLR-db1, WMLR-db2, WMLR-db3, WMLR-db8 and single MLR and ANN models. During testing period, the values of determination coefficient (R2) and RMSE for WMLR-db10 model for two-, four-, seven- and 14-day lead time were found to be, respectively, 0.996 (751.87 m3·s–1), 0.991 (1,174.80 m3·s–1), 0.984 (1,585.02 m3·s–1), and 0.968 (2,196.46 m3·s–1). Also, it was observed that for lower order wavelets (db1, db2, db3) WANN’s performance was better, and for higher order wavelets (db8, db10) WMLR’s performance was better. Correspondingly, it was observed that all hybrid models’ efficiency increased with increase in the decomposition level.

Deep learning CNN-APSO-LSSVM hybrid fusion model for feature optimization and gas-bearing prediction

Conventional machine learning (CML) methods have been successfully applied for gas reservoir prediction. Their prediction accuracy largely depends on the quality of the sample data; therefore, feature optimization of the input samples is particularly important. Commonly used feature optimization methods increase the interpretability of gas reservoirs; however, their steps are cumbersome, and the selected features cannot sufficiently guide CML models to mine the intrinsic features of sample data efficiently. In contrast to CML methods, deep learning (DL) methods can directly extract the important features of targets from raw data. Therefore, this study proposes a feature optimization and gas-bearing prediction method based on a hybrid fusion model that combines a convolutional neural network (CNN) and an adaptive particle swarm optimization-least squares support vector machine (APSO-LSSVM). This model adopts an end-to-end algorithm structure to directly extract features from sensitive multicomponent seismic attributes, considerably simplifying the feature optimization. A CNN was used for feature optimization to highlight sensitive gas reservoir information. APSO-LSSVM was used to fully learn the relationship between the features extracted by the CNN to obtain the prediction results. The constructed hybrid fusion model improves gas-bearing prediction accuracy through two processes of feature optimization and intelligent prediction, giving full play to the advantages of DL and CML methods. The prediction results obtained are better than those of a single CNN model or APSO-LSSVM model. In the feature optimization process of multicomponent seismic attribute data, CNN has demonstrated better gas reservoir feature extraction capabilities than commonly used attribute optimization methods. In the prediction process, the APSO-LSSVM model can learn the gas reservoir characteristics better than the LSSVM model and has a higher prediction accuracy. The constructed CNN-APSO-LSSVM model had lower errors and a better fit on the test dataset than the other individual models. This method proves the effectiveness of DL technology for the feature extraction of gas reservoirs and provides a feasible way to combine DL and CML technologies to predict gas reservoirs.

Toward Improving the Generation Quality of Autoregressive Slot VAEs.

Unconditional scene inference and generation are challenging to learn jointly with a single compositional model. Despite encouraging progress on models that extract object-centric representations ("slots") from images, unconditional generation of scenes from slots has received less attention. This is primarily because learning the multiobject relations necessary to imagine coherent scenes is difficult. We hypothesize that most existing slot-based models have a limited ability to learn object correlations. We propose two improvements that strengthen object correlation learning. The first is to condition the slots on a global, scene-level variable that captures higher-order correlations between slots. Second, we address the fundamental lack of a canonical order for objects in images by proposing to learn a consistent order to use for the autoregressive generation of scene objects. Specifically, we train an autoregressive slot prior to sequentially generate scene objects following a learned order. Ordered slot inference entails first estimating a randomly ordered set of slots using existing approaches for extracting slots from images, then aligning those slots to ordered slots generated autoregressively with the slot prior. Our experiments across three multiobject environments demonstrate clear gains in unconditional scene generation quality. Detailed ablation studies are also provided that validate the two proposed improvements.

Tailoring of physical, optical, and thermal properties of Se50-xTe30Ge20Sbx chalcogenide glasses: Influence of metalloids

The compositional dependence on physical and spectroscopic characteristics of chalcogenide Se50-xTe30Ge20Sbx (x = 5, 10, 15, and 20 at. wt%) bulk glassy systems has been explored in this communication. Numerous physical and optical features are rigorously examined by using the best glass sample among all the prepared samples. The XRD pattern reveals the amorphous character of the synthesized materials. Some of the physical parameters like density, and compactness are gradually increasing with the Sb addition, while the free volume percentage and excess volume follow the decreasing trend. The optical properties have been precisely inspected by UV-Vis spectroscopy in the (200–800) nm spectral range. The optical bandgap (Eopt) values are found to decline from (1.09–0.85) eV with Sb content. The observed refractive index, n values experience a subsequent decline with incident energy. By using the Wemple and DiDomenico (WDD) single oscillator model, the refractive index dispersion curve has been studied, and the dispersion parameters are discussed. The valence band (VB) and conduction band (CB) locations are estimated, and the obtained result reveals the VB potential increases from −15,089 eV to −1.3899 eV, while that of CB decreases from −0.4113 eV to −0.5303 eV with the rise of Sb-concentration. The cohesive energy (CE) values have been estimated by using the chemical bond approach (CBA). Investigation reveals the CE of the system decreases from 45.237 kcal/mole to 44.229 kcal/mole justifying the decrement in obtained band gap energy values of the studied glassy systems. Furthermore, the compositional dependency of various thermal parameters is investigated by using the DSC approach. The glass transition (Tg), and peak crystallization (Tp) were significantly affected by Sb addition. The obtained values of Tg are found to decrease from (337–305 K), while the thermal stability is found to increase with the Sb incorporation.

Identifying failure types in cyber-physical water distribution networks using machine learning models

Abstract Water cyber-physical systems (CPSs) have experienced anomalies from cyber-physical attacks as well as conventional physical and operational failures (e.g., pipe leaks/bursts). In this regard, rapidly distinguishing and identifying a facing failure event from other possible failure events is necessary to take rapid emergency and recovery actions and, in turn, strengthen system's resilience. This paper investigated the performance of machine learning classification models – support vector machine (SVM), random forest (RF), and artificial neural networks (ANNs) – to differentiate and identify failure events that can occur in a water distribution network (WDN). Datasets for model features related to tank water levels, nodal pressure, and water flow of pumps and valves were produced using hydraulic model simulation (WNTR and epanetCPA tools) for C-Town WDN under pipe leaks/bursts, cyber-attacks, and physical attacks. The evaluation of accuracy, precision, recall, and F1-score for the three models in failure type identification showed the variation of their performances depending on the specific failure types and data noise levels. Based on the findings, this study discussed insights into building a framework consisting of multiple classification models, rather than relying on a single best-performing model, for the reliable classification and identification of failure types in WDNs.

Long-term ozone exposure and mortality in patients with chronic kidney disease: a large cohort study

BackgroundEpidemiologic studies on the effects of long-term exposure to ozone (O3) have shown inconclusive results. It is unclear whether to O3 has an effect on chronic kidney disease (CKD). We investigated the effects of O3 on mortality and renal outcome in CKD.MethodsWe included 61,073 participants and applied Cox proportional hazards models to examine the effects of ozone on the risk of end-stage renal disease (ESRD) and mortality in a two-pollutants model adjusted for socioeconomic status. We calculated the concentration of ozone exposure one year before enrollment and used inverse distance weighting (IDW) for interpolation, where the exposure was evenly distributed.ResultsIn the single pollutant model, O3 was significantly associated with an increased risk of ESRD and all-cause mortality. Based on the O3 concentration from IDW interpolation, this moving O3 average was significantly associated with an increased risk of ESRD and all-cause mortality. In a two-pollutants model, even after we adjusted for other measured pollutants, nitrogen dioxide did not attenuate the result for O3. The hazard ratio (HR) value for the district-level assessment is 1.025 with a 95% confidence interval (CI) of 1.014–1.035, while for the point-level assessment, the HR value is 1.04 with a 95% CI of 1.035–1.045. The impact of ozone on ESRD, hazard ratio (HR) values are, 1.049(95%CI: 1.044–1.054) at the district unit and 1.04 (95%CI: 1.031–1.05) at the individual address of the exposure assessment. The ozone hazard ratio for all-cause mortality was 1.012 (95% confidence interval: 1.008–1.017) for administrative districts and 1.04 (95% confidence interval: 1.031–1.05) for individual addresses.ConclusionsThis study suggests that long-term ambient O3 increases the risk of ESRD and mortality in CKD. The strategy to decrease O3 emissions will substantially benefit health and the environment.

Explainable ensemble machine learning revealing the effect of meteorology and sources on ozone formation in megacity Hangzhou, China

Megacity Hangzhou, located in eastern China, has experienced severe O3 pollution in recent years, thereby clarifying the key drivers of the formation is essential to suppress O3 deterioration. In this study, the ensemble machine learning model (EML) coupled with Shapley additive explanations (SHAP), and positive matrix factorization were used to explore the impact of various factors (including meteorology, chemical components, sources) on O3 formation during the whole period, pollution days, and typical persistent pollution events from April to October in 2021–2022. The EML model achieved better performance than the single model, with R2 values of 0.91. SHAP analysis revealed that meteorological conditions had the greatest effects on O3 variability with the contribution of 57 %–60 % for different pollution levels, and the main drivers were relative humidity and radiation. The effects of chemical factors on O3 formation presented a positive response to volatile organic compounds (VOCs) and fine particulate matter (PM2.5), and a negative response to nitrogen oxides (NOx). Oxygenated compounds (OVOCs), alkenes, and aromatic of VOCs subgroups had higher contribution; additionally, the effects of PM2.5 and NOx were also important and increased with the O3 deterioration. The impact of seven emission sources on O3 formation in Hangzhou indicated that vehicle exhaust (35 %), biomass combustion (16 %), and biogenic emissions (12 %) were the dominant drivers. However, for the O3 pollution days, the effects of biomass combustion and biogenic emissions increased. Especially in persistent pollution events with highest O3 concentrations, the magnitude of biogenic emission effect elevated significantly by 156 % compared to the whole situations. Our finding revealed that the combination of the EML model and SHAP analysis could provide a reliable method for rapid diagnosis of the cause of O3 pollution at different event scales, supporting the formulation of control measures.