As electricity market reforms deepen, the electricity retail sector faces multifaceted risks, including price volatility, customer defaults, and the integration of renewable energy sources. These challenges necessitate a robust risk assessment framework. Effective risk management is particularly critical in the textile industry, where fluctuations in energy consumption costs significantly affect operational budgeting and production stability. This study proposes a risk assessment framework based on the Gradient Boosting Decision Tree (GBDT) algorithm, leveraging feature importance analysis to identify key indicators and integrating an adaptive weighted loss function to enhance the detection of highrisk events. Experiments conducted on 35,000 transaction records from a provincial electricity market demonstrate that the proposed model achieves an F1 score of 0.89, representing improvements of 21.4% over Logistic Regression (LR) (F1 = 0.73) and 9.8% over Random Forest (RF). In addition, the proposed model attains an area under the curve (AUC) of 0.94, representing improvements of 21.9% and 6.8% over LR and RF models, respectively. The prediction accuracy for high-risk events, such as payment arrears and sudden load drops, reaches 91.5%. The results indicate that the proposed method significantly enhances model robustness through ensemble learning and provides data-driven decision support for electricity retailers to develop differentiated risk management strategies. The effectiveness of advanced machine learning algorithms in assessing risks in complex energy markets is thus verified. Similarly, these algorithms are invaluable in the textile industry for assessing the complex risks associated with raw material price volatility and demand forecasting, enabling manufacturers to optimize inventory and production planning.

electricity retail market, risk assessment, gradient boosting decision tree, machine learning