China's crude oil futures present pronounced nonlinearities, time-varying dynamics, and inherent uncertainty, making accurate volatility forecasting essential for effective risk management. However, most of the forecasting models face a fundamental trade-off: linear HAR models lack nonlinear flexibility, while mainstream machine learning methods operate as black boxes with limited interpretability. Meanwhile, conventional fuzzy inference systems, e.g., ANFIS, offer rule-based transparency but suffer from semantic inconsistency with different membership functions for the same linguistic value produce varying reasoning outcomes. To overcome these limitations, we develop a parsimonious and interpretable realized volatility forecasting model, RV-LCW-FIS, which based on labeled computing-with-words and fuzzy inference system. The key innovation lies in a labeling mechanism that anchors linguistic values to intervals derived from Gaussian-sum density clustering, ensuring consistent semantic interpretation while preserving parsimony. The model mines interpretable fuzzy linguistic If-Then rules from daily, weekly, and monthly volatility components and employs an improved compositional rule of inference for forecasting. Using high-frequency data from China's crude oil futures market, out-of-sample tests show that the RV-LCW-FIS model outperforms HAR-RV, HAR-RV-CJ, as well as SVM, RF, XGBoost, RNN, MLP, LSTM, and ANFIS. The model generates a fuzzy linguistic knowledge base, identifying four distinct volatility transmission patterns and linking predictive accuracy with practical interpretability for risk management.

Data missing and noise problems are often encountered when predicting financial distress in real-world scenarios. To address and eliminate the negative effects of missing and noisy data, a novel case-based reasoning (CBR)-driven clustering imputation and noise-resistant (ClusImpute-NoisRes) classification learning paradigm is proposed for financial distress prediction to achieve excellent imputation and prediction performance. In this learning paradigm, CBR-driven clustering imputation and CBR-driven noise-resistant classifier prediction are two primary stages. In the first stage, a clustering-based hybrid CBR-driven weighted (ClusHyCBR) imputation method is introduced to handle the issue of missing data and their uneven distribution. In the second stage, a CBR-driven noise-resistant classification model is constructed to identify class noise and reduce the negative interference of class noise on the prediction model. For illustration and verification, a dataset of Chinese-listed enterprises and its derived multiple datasets with different missing degrees and noise levels are used to conduct the experimental study. Experimental results demonstrate that the proposed ClusHyCBR imputation method consistently outperforms competing methods, improving Type II accuracy by 1.92 %-8.99 % on the original dataset, with increasingly larger gains on higher missing degrees. The proposed CBR-driven noise-resistant classification model maintains noise identification accuracy above 0.8722 and Type II accuracy above 0.7022 after injecting 10 %-50 % class noise, which is significantly higher than that of the base classifier. These outcomes indicate that the CBR-driven ClusImpute-NoisRes classification learning paradigm provides a viable solution for enterprises, regulatory and policy-making bodies, and market participants to support prediction and warning of financial distress with missing and noisy data.

Class imbalance and class overlapping are two prevalent data traits that collectively impair the generalisation performance of credit classification models. For this superposition trait, conventional sampling techniques or remedial measures either amplify marginal noise or irreversibly discard useful majority information. To address these issues, a hybrid sampling method integrating a decision tree (DT) and a subregional synthetic minority oversampling technique (SR-SMOTE) algorithm is proposed. On the one hand, a decision tree-based undersampling (DTU) algorithm is proposed, which innovatively embeds classifier training into the sampling process to iteratively identify and eliminate hard-to-classify samples from the majority class. This can effectively retain the key information of the majority class and alleviate the problem of data imbalance. On the other hand, an adaptive SR-SMOTE algorithm is developed, which partitions the minority class boundary into subregions based on local class overlap density and performs weighted oversampling within each subregion. This focuses on the most ambiguous areas and mitigates the risk of generating overlapping synthetic samples, thereby reducing inter-class overlap in imbalanced data. The proposed method is validated on four imbalanced credit datasets. Experimental results show that it outperforms benchmark models. Robustness tests with GBDT, KNN, and LogR confirm consistent performance gains, with GBDT achieving the best results. This suggests the proposed approach is effective for credit classification with imbalanced and overlapping data.

Detecting credit card fraud in streaming data demands simultaneous adaptation to class imbalance and concept drift—two challenges that degrade model performance. A unified Adaptive Online Bagging with Hybrid Sampling (AOBHS) framework is designed to address both challenges in streaming fraud detection. AOBHS employs a drift-sensitive sketch estimator to guide hybrid sampling of underrepresented fraudulent cases, generating temporally relevant minority samples to preserve class balance amid concept drifts. Furthermore, by combining active drift detection with passive performance-based replacement of underperforming classifiers, AOBHS adopts a dual-mode drift adaptation strategy that addresses both abrupt and gradual drifts, enhancing fraud identification accuracy over time. Comprehensive empirical evaluations across 108 synthetic datasets with 9 drift types, 5 real-world datasets across multiple domains, and a credit card transactions case study, demonstrate that AOBHS maintains robust and competitive performance compared to established benchmarks, with particular strengths in Recall, G-mean, and operational cost. These results confirm that AOBHS is particularly suited for scenarios where high recall and low operational cost are critical, and reveal its practical value for strengthening financial risk management and operational resilience in digital payment systems.

Financial fraud detection is crucial in the banking and financial sectors, but it faces considerable challenges because of class imbalance in transaction data and the growing threat of adversarial attacks. These issues frequently undermine the effectiveness of deep neural networks despite their demonstrated potential in this domain. To address these challenges, this paper proposes a novel approach, adversarial training–based deep imbalanced learning (ATDIL), which integrates imbalanced learning and adversarial defense into a unified approach. ATDIL leverages an adversarial autoencoder to efficiently synthesize high-quality minority-class samples that are informative and adversarial, while maintaining low computational complexity. The effectiveness of ATDIL is rigorously validated through both theoretical and experimental analyses. Theoretically, the optimal solution form for the inner optimization problem in ATDIL is derived, and its convergence under mild assumptions is established. Extensive evaluations on seven real-world financial data sets demonstrate that ATDIL outperforms state-of-the-art imbalanced learning methods across multiple metrics while exhibiting superior resilience under adversarial conditions. This combination of theoretical guarantees and empirical evidence highlights ATDIL’s ability to effectively address class imbalance and model security, offering a robust and practical framework for enhancing financial fraud detection systems. History: Accepted by Ram Ramesh, Area Editor for Data Science and Machine Learning. Funding: This work was supported by the National Natural Science Foundation of China [Grant 72401208], the Key Program of the National Natural Science Foundation of China [Grant 72331007], the Natural Science Foundation of Sichuan Province [Grant 2025NSFSC1981], the Postdoctoral Fellowship Program of the China Postdoctoral Science Foundation [Grant GZB20240504], the International Visiting Program for Excellent Young Scholars of Sichuan University (SCU), and the Humanities and Social Science Youth Foundation of the Ministry of Education of China [Grant 23YJCZH088]. Supplemental Material: The software that supports the findings of this study is available within the paper and its Supplemental Information ( https://pubsonline.informs.org/doi/suppl/10.1287/ijoc.2025.1251 ) as well as from the IJOC GitHub software repository ( https://github.com/INFORMSJoC/2025.1251 ). The complete IJOC Software and Data Repository is available at https://informsjoc.github.io/ .

We construct a text-based measure of oil market uncertainty using Wall Street Journal front-page articles from January 1986 to December 2022. This measure, the news implied oil volatility (NOVX), is derived from over 100,000 articles and reflects oil market uncertainty during key historical events. NOVX emerges as a significant predictor of oil returns, both in-sample and out-of-sample, outperforming existing predictors like real-time, interest rate, and macroeconomic variables, as well as other news-based indices. Following Manela and Moreira (2017), we decompose NOVX to identify different rare disaster concerns. Among these concerns, those related to government and natural disasters play particularly significant roles in forecasting oil returns. Additionally, we identify an economic mechanism: increases in news-implied oil volatility reduce oil production and economic activity, while increasing oil inventories and decreasing oil prices.

This article proposes a two-phase unilateral rule extraction method based on a multiobjective genetic algorithm (URE-MOGA) to solve the challenges of sparse categorical features and model interpretability in credit risk classification tasks. In this method, rule encoding is employed to represent the sparse categorical features without increasing data dimensionality. The Pareto dominance theory is utilized to balance the tradeoff between model accuracy and interpretability. Additionally, a diversity function (DF) and unilateral rule learning strategy are designed to facilitate the acquisition of reasonable rules and classifiers by the URE-MOGA model. Furthermore, three publicly available datasets, along with six derived datasets, are used to validate the feasibility of the proposed URE-MOGA model. Experimental results demonstrate that the proposed model outperforms baseline methods in terms of accuracy, interpretability, and robustness. The adoption of a multiobjective approach enables the URE-MOGA model to extract more concise and comprehensible classification rules while ensuring high accuracy levels. Moreover, both the unilateral rule learning strategy and DF play vital roles in enhancing classifier accuracy and robustness within the URE-MOGA model. Overall, the proposed URE-MOGA model provides a novel insight into categorical feature sparsity handling interpretable modeling for credit risk classification tasks.

The global power system is currently undergoing three fundamental transitions: decarbonization, marketization, and digitalization. These changes have transcended merely technical considerations, positioning themselves as key arenas for geopolitical competition and socio-economic restructuring. A complementary and coordinated development model among diverse power generation agents can effectively reconcile the seemingly contradictory goals of transitioning to low-carbon energy and ensuring energy security. Our study leverages Empirical Mode Decomposition to establish an analytical framework for China's power generation coordination. Resource endowments, economic growth, market structure, and climatic and environmental conditions core factors influencing the coordinated development of power generation, with market structure superseding resources as the pivotal variable. By 2030, the share of renewable energy in each Chinese province will reach 23

Travel mode identification (TMI) is crucial for intelligent transportation systems but still lacks sufficient applicability in real-world scenarios due to trajectory data noise. Existing noise handling approaches often rely on simple filtering and smoothing techniques, or on random perturbations for noise augmentation; however, both strategies may obscure key behavioral patterns and risk data leakage during noise filtering (i.e., when mode-specific filtering is applied to test data, implicitly using label information). This paper proposes a novel noise-robust TMI framework. First, a hybrid preprocessing strategy is introduced that preserves both clean and noisy data for robust representation learning and to prevent data leakage by testing on real-world noise. Second, a behavior-indication mask method is developed that identifies critical behavioral change points in both trajectory segments and motion features to guide strategic noise injection during model training. Third, a dual-branch transformer architecture is constructed that processes trajectory and motion features that simultaneously leverage spatial and motion feature semantics. A two-stage training strategy is designed: hybrid data and mask-guided noise injection pretraining for enhanced noise robustness, followed by feature fusion fine-tuning to integrate complementary information, thereby ultimately improving TMI performance and noise robustness. Extensive experiments on two real-world datasets demonstrate that our method consistently outperforms baseline models, exhibiting greater stability as the noise ratio increases. Additional generalization analysis under diverse simulated noise scenarios further validates the model’s robustness against out-of-distribution noise patterns. These findings indicate the effectiveness of our approach in practical application scenarios with real-world noise.

The exponential growth of the investable universe poses a significant computational challenge for large-scale portfolio selection, rendering traditional optimization algorithms inefficient. To address this challenge, a coherent ising machine (CIM) based hybrid quantum-classical framework is proposed to optimize the asset preselection problem for multi-asset-class portfolios. In this framework, the asset preselection is first formulated as a distance maximization problem and then transformed into a quadratic unconstrained binary optimization (QUBO) form. Finally, a hybrid quantum-classical strategy is employed to address both the NP-hard structure of the problem and the precision limitations of current CIM hardware, enabling the identification of a diversified preselected asset pool via CIM for subsequent portfolio allocation. Empirical results on U.S. stocks and global commodity futures show that the proposed framework constructs diversified multi-asset-class pools and delivers strong out-of-sample performance. It solves the baseline instances at millisecond-level latency and achieves a favorable solution-quality-runtime tradeoff relative to the benchmark algorithms. These findings demonstrate the practical potential of CIM-based optimization for large-scale portfolio preselection.

Evolutionary feature construction has shown strong promise in symbolic regression by automatically discovering informative transformations of input features that enhance a simple base learner. However, existing approaches often lack explicit mechanisms to preserve important constructed features discovered during evolution, and valuable genetic material can be lost when genetic operators disrupt effective features. This paper introduces an adaptive protection mechanism that leverages feature importance metrics to selectively preserve constructed features during evolution. The mechanism provides stronger protection for more important constructed features while still allowing less important features to be modified and to incorporate useful building blocks from more important features. We evaluate the approach using multiple feature importance calculation methods and demonstrate its robustness across different base learners. Experimental results on 98 regression benchmark datasets show that the proposed mechanism consistently improves solution quality over baseline approaches, and experiments on two credit classification datasets demonstrate that the method also extends effectively to improve search effectiveness beyond symbolic regression.

Natural gas demand forecasting is challenged by pattern heterogeneity and recurring cycles, which existing global modeling approaches cannot effectively address. This study proposes a latent pattern retrieval and expert specialization (LPRES) framework that strategically leverages pretrained language models (PLMs) not as direct forecasting tools but through role-specific adaptation. First, latent pattern awareness is developed through a temporal feature learning model derived from a PLM fine-tuned for anomaly detection, exploiting its sensitivity to pattern changes. Second, based on this model, an adaptive sliding window segmentation algorithm partitions historical data into segments, each corresponding to a distinct latent pattern. Third, for each identified latent pattern, specialized forecasting experts are trained using a PLM fine-tuned on large-scale forecasting tasks, thereby adapting its strong predictive capacity to the characteristics of individual latent patterns. Fourth, during forecasting, input windows are matched to their most similar latent pattern through similarity-based retrieval and routed to the corresponding expert. Experiments on four natural gas datasets spanning monthly and hourly frequencies show that LPRES delivers competitive forecasting performance across diverse data characteristics, achieving mean absolute percentage error (MAPE) values from 0.035 to 0.107 and reducing errors by up to 11.2% relative to the strongest baselines. A complementary theoretical framework identifies the conditions under which expert specialization is most beneficial.

ABSTRACT In the field of credit risk classification, the prediction model needs to meet a large number of labelled samples to make the model fully trained. However, in some practical scenarios, it is difficult for financial institutions to obtain a large number of available labelled samples because of the cost or privacy. Shortage of samples will cause the prediction model ineffective or even invalid. To address low sample size problem, a novel similarity‐based graph neural network with meta‐learning called ML‐SGNN is proposed for credit classification with low sample size. In the proposed ML‐SGNN methodology, attribute co‐training based on meta‐learning strategy is first introduced to increase the generalisation capability of the proposed model under label scarcity condition. Then, a node update strategy based on similarity is proposed to aggregate features of samples with same label so as to provide a contrastive signal that increases the discriminative information extracted from low sample size. Finally, a novel similarity‐based edge update strategy utilising the difference of node features and convolutional neural network is proposed to better describe the similarity relationships among unordered attribute structures to improve the generalizability of ML‐SGNN. The experiment results demonstrate that the proposed ML‐SGNN method consistently achieves the highest average ranking across different sample sizes and evaluation metrics among the compared methods. Generalisation capability of ML‐SGNN indicating that ML‐SGNN can be considered as a promising solution for credit classification with low sample size.

Building an intelligent diagnosis model based on clinical data improves the efficiency of medical decision-making and reduces healthcare costs. However, the uneven distribution of disease prevalence (imbalanced dataset) and multiple clinical features (high-dimensional features) significantly impact the model's prediction performance, leading to occurrences of misdiagnosis and underdiagnosis. This paper proposes an Adaptive Cost-sensitive Neural Network with Metric Pre-training (AdaCSNN-MPT) to address the above issues. Firstly, metric pre-training is pro posed to learn effective low-dimensional representations and provide better parameter initialization. Secondly, an adaptive cost-sensitive neural network is proposed to improve the overall classification performance by iden tifying hard-to-classify invaded samples. Experiments on six real-world disease diagnosis tasks demonstrate that AdaCSNN-MPT outperforms the baseline models.

The global climate issue has driven carbon reduction actions, but the unequal emission reduction policies have led to competitiveness loss and carbon leakage. Carbon tariffs are regarded as one of the solutions despite the controversies. Given developed countries' economic status and their determination to levy carbon tariffs, as well as the crucial role China's energy and power industries play in energy supply, the effects of carbon tariffs on these industries necessitate further investigation. This paper proposed a theoretical analysis mechanism to illustrate how the economy, trade, industrial price, industrial output, and industrial emissions are affected by carbon tariffs and utilized the Energy-Environmental Version of the Global Trade Analysis Project model (GTAP-E) to simulate seven scenarios, providing quantitative support for the theoretical analysis. The findings show that: (1) Carbon tariff policy is ineffective in ensuring domestic product competitiveness and mitigating the carbon leakage problem. It fails to significantly improve the economy of the levying countries. Moreover, it has a negligible effect on emissions reduction in the affected countries and will affect their economy and exports. (2) The proactive carbon tax in China will neither damage the emission reduction achievements of the EU, US and Japan nor cause a growth in global emissions. As an effective response, it can fundamentally change the energy use and production mode, achieve a good emission reduction effect, and mitigate the adverse impact from carbon tariffs. (3) The differentiated carbon taxes, which comprehensively account for the cost fairness among countries, contribute more effectively to alleviating the transformation pressure on China's energy and power industries. The paper assesses how reactive carbon tariffs and proactive carbon taxes affect China's energy and power industries, which differentiates from previous studies and can provide a completely new analytical perspective for evaluating the effect of carbon tariffs.

International Journal of Information Technology & Decision MakingAccepted Papers No AccessForecasting Crude Oil Price with Embedding Convolutional Neural Network FrameworkKaijian He, Lean Yu, Jia Liu, and Yingchao ZouKaijian He, Lean Yu Search for more papers by this author , Jia Liu Search for more papers by this author , and Yingchao Zouhttps://orcid.org/0000-0003-3965-4776 Search for more papers by this author https://doi.org/10.1142/S0219622025410020Cited by:0 (Source: Crossref) PreviousNext AboutFiguresReferencesRelatedDetailsPDF/EPUB ToolsAdd to favoritesDownload CitationsTrack CitationsRecommend to Library ShareShare onFacebookTwitterLinked InRedditEmail Cite Recommend We recommendA Cooperative Attack Detection Framework for MANET-IOT Network using Optimized Gradient Boosting Convolutional Neural NetworkP Sathyaraj, Journal of Circuits, Systems and Computers, 2023NEURAL NETWORKS FOR THE OPTIMIZATION OF CRUDE OIL BLENDINGInternational Journal of Neural Systems, 2011A Convolutional Neural Network Image Compression Algorithm for UAVsYongdong Dai, Journal of Circuits, Systems and Computers, 2024Optimizing Convolutional Neural Network Accelerator on Low-Cost FPGATruong Quang Vinh, Journal of Circuits, Systems and Computers, 2021High Performance Kernel Architecture for Convolutional Neural Network AccelerationAnakhi Hazarika, Journal of Circuits, Systems and Computers, 2021A Temporal Convolutional Network Based Hybrid Model for Short-term Electricity Price Forecasting Haoran Zhang, CSEE Journal of Power and Energy Systems, 2024Residential Appliance Detection Using Attention- based Deep Convolutional Neural Network Chunyu Deng, CSEE Journal of Power and Energy Systems, 2022Power Grid Fault Diagnosis Based on Deep Pyramid Convolutional Neural Network Xu Zhang, CSEE Journal of Power and Energy Systems, 2023Detecting Double Mixed Compressed Images Based on Quaternion Convolutional Neural Network WANG Hao, Chinese Journal of Electronics, 2023A causal convolutional neural network for multi-subject motion modeling and generation Shuaiying Hou, Computational Visual Media, 2023Powered by Privacy policyGoogle Analytics settings FiguresReferencesRelatedDetailsNone Recommended Recommended FORECASTING THE CRUDE OIL SPOT PRICE BY WAVELET NEURAL NETWORKS USING OECD PETROLEUM INVENTORY LEVELSYE PANG, WEI XU, LEAN YU, JIAN MA, KIN KEUNG LAI, SHOUYANG WANG, and SHANYING XUNew Mathematics and Natural ComputationVol. 07, No. 02A DEVELOPED WAVELET-BASED LOCAL LINEAR NEURO FUZZY MODEL FOR THE FORECASTING OF CRUDE OIL PRICEHOSSEIN IRANMANESH, MAJID ABDOLLAHZADE, ARASH MIRANIAN, and HOSSEIN HASSANIInternational Journal of Energy and StatisticsVol. 01, No. 03THE MORE THE BETTER: FORECASTING OIL PRICE WITH DECOMPOSITION-BASED VECTOR AUTOREGRESSIVE MODELHAIBIN XIE, XUN ZHANG, and SHOUYANG WANGInternational Journal of Energy and StatisticsVol. 01, No. 01Predicting Crude Oil Future Price Using Traditional and Artificial Intelligence-Based Model: Comparative AnalysisSanjeev Kadam, Anshul Agrawal, Aryan Bajaj, Rachit Agarwal, Rameesha Kalra, and Jaymin ShahJournal of International Commerce, Economics and PolicyVol. 14, No. 03Spatial Decomposition and Aggregation for Attention in Convolutional Neural NetworksMeng Zhu, Weidong Min, Hongyue Xiang, Cheng Zha, Zheng Huang, Longfei Li, and Qiyan FuInternational Journal of Pattern Recognition and Artificial IntelligenceVol. 38, No. 01Forecasting crude oil price with ensemble neural networks based on different feature subsets methodAli Moosavi, Seyyed Hossein Khasteh, and Mohammad Ali BagheriInternational Journal of Energy and StatisticsVol. 03, No. 02U.S. DIESEL FUEL PRICE RESPONSES TO THE GLOBAL CRUDE OIL SUPPLY AND DEMANDBAHRAM ADRANGI, ARJUN CHATRATH, JOSEPH MACRI, and KAMBIZ RAFFIEEAnnals of Financial EconomicsVol. 13, No. 04AN INTEGRATED MODEL USING WAVELET DECOMPOSITION AND LEAST SQUARES SUPPORT VECTOR MACHINES FOR MONTHLY CRUDE OIL PRICES FORECASTINGYEJING BAO, XUN ZHANG, LEAN YU, KIN KEUNG LAI, and SHOUYANG WANGNew Mathematics and Natural ComputationVol. 07, No. 02 Accepted Papers Metrics Downloaded 0 times History Received 24 April 2023 Accepted 24 December 2024 PDF download

The prevalent challenge of class sparsity issues in credit risk classification commonly focuses on instance-view solutions, while feature-view solutions are overlooked. For this purpose, this paper designs a dual-view ensemble learning model to tackle class sparsity and its associated traits of overlap, noise, and irrelevance. The model comprises two phases integrated into a recurrent structure. Firstly, an instance-view dynamic sampling method is developed on instance importance estimation to select important instances. Secondly, at the feature view, a feature fusion network is introduced to extract classification features by integrating feature interaction and densely connected structures. In order to form a recurrent structure, the trained network serves as an instance importance estimator in the subsequent epoch. The proposed model is evaluated using four publicly available datasets and six derived datasets, and experimental results demonstrate its excellent performance relative to other benchmarks. This indicates the proposed ensemble model presents an effective and competitive solution for credit risk classification in scenarios with class sparsity.

To solve the high-dimensional issue in credit risk assessment, a hybrid clustering and boosting tree feature selection method is proposed. In the hybrid methodology, an improved minimum spanning tree model is first used to remove redundant and irrelevant features. Then three embedded feature selection approaches (i.e., Random Forest, XGBoost, and AdaBoost) are used to further enhance the feature-ranking efficiency and obtain better prediction performance by applying the optimal features. For verification purpose, two real-world credit datasets are used to demonstrate the effectiveness of the proposed hybrid clustering and boosting tree feature selection (CBTFS) methodology. Experimental results demonstrated that the proposed method is superior to others classic feature selection methods. This indicates that the proposed hybrid clustering and boosting tree feature selection method can be used as a promising tool for solving high-dimensional issue in credit risk assessment. First published online 12 February 2025

The Renewable Portfolio Standards (RPS) policy imposes mandatory obligations on the obligated entities for the consumption of renewable energy electricity (RE), which can be completed through RE consumption in the electricity market, participating in Tradeable Green Certificates (TGC) transactions and consumption above quota (CAQ) transactions in supplementary markets. In this context, how to coordinate the three compliance options and how to design the RPS policy become key issues. This paper proposes a hybrid agent-based model (ABM) for RPS by incorporating an evolutionary training algorithm to depict the micro-level compliance behaviors and decision-making process of obligated entities and evaluate the macro-level impacts of various RPS targets. Our analysis reveals that 1) RPS target can positively promote the RE electricity consumption, but the trend of change in CAQ consumption share and consumers' profit with the RPS target regulation is nonmonotonic; 2) the CAQ market and TGC market exhibit a positive price linkage, tightening RPS constraints can significantly stimulate the vitality of both markets; 3) when RPS target is set at 60 %, the total consumption obligation and actual RE consumption exceed the RPS target the most; when set at 90 %, the gap between actual RE consumption and RPS target is the largest.