With the rapid expansion of e-commerce, last-mile delivery in express logistics faces significant challenges, including low efficiency and high operational costs. Taking the Xiqing District of Tianjin as a case study, this research proposes a three-stage framework integrating complex network theory and machine learning. First, the Louvain algorithm is employed to achieve intelligent partitioning of delivery areas, resulting in a modularity increase to 0.789. Second, an eXtreme Gradient Boosting (XGBoost) model is utilized to predict terminal service modes, achieving an accuracy of 87.8%. Finally, a route planning model is constructed using Particle Swarm Optimization (PSO). To validate these methods, a three-day logistics system simulation was conducted via AnyLogic to evaluate the effectiveness of different delivery policies. The results demonstrate that, compared to traditional independent delivery, the joint delivery approach reduces total costs by 25.32%. Furthermore, by introducing a carbon emission accounting model, leading to an estimated 25% reduction in daily carbon emissions, achieving a win-win situation for both economic and environmental benefits.

The increasing integration of environmental, social, and governance (ESG) considerations into financial markets has raised fundamental questions regarding their roles in investment decision making. In particular, it remains unclear whether ESG-oriented investment strategies mirror substantive changes in portfolio construction or primarily follow prevailing market trends. This study examined the decision relevance of ESG signals by analyzing the behavior and performance of ESG-oriented mutual funds. Using a sample of 41 funds, a data-driven analytical framework was employed to evaluate portfolio composition, risk–return performance, and the determinants of ESG ratings. The analysis first considered whether ESG funds systematically allocated capital toward firms with stronger ESG profiles. Although a modest tilt toward higher-rated companies was observed, the differences relative to conventional funds remained limited and, in most cases, statistically insignificant, thus indicating that ESG considerations were not the dominant driver of portfolio selection. The second part evaluated fund performance within a risk–return framework using benchmark-adjusted measures and information ratios (IR) . While all funds generated positive returns over the sample period, the majority failed to outperform their benchmarks. Only a small subset exhibited consistently favorable risk-adjusted performance. These findings suggested that ESG-oriented strategies did not provide a reliable basis for achieving superior financial outcomes and might involve trade-offs in portfolio allocation. Finally, cross-sectional regression analysis demonstrated that ESG ratings were strongly associated with firm-specific characteristics, especially size and profitability. This result indicated that ESG scores might reflect underlying financial capacity rather than deliberate sustainability-oriented decisions. Taken together, this study implied that ESG signals offered limited standalone values for guiding decisions of investment. Effective portfolio design therefore requires a broader analytical approach that integrates ESG metrics with conventional financial indicators.

Customer retention in the telecommunications industry poses a critical challenge in data-driven business operations, while high predictive accuracy does not necessarily translate into superior commercial outcomes under asymmetric misclassification costs. This study investigated a profit-oriented decision analytics framework for customer churn management by integrating predictive performance with business value optimization. A cost-sensitive Cost-Sensitive Improved Sparrow Search Strategy Algorithm Stacking (CS-ISSA-Stacking) framework was developed by incorporating customer lifetime value (CLV) and marketing intervention costs into an Expected Total Profit (ETP) objective function. An Improved Sparrow Search Algorithm (ISSA) was constructed using Tent chaotic mapping and Cauchy mutation mechanisms to enhance global optimization capability. EXtreme Gradient Boosting (XGBoost), Light Gradient Boosting Machine (LightGBM), and Categorical Boosting (CatBoost) were employed as heterogeneous base learners, and ISSA was used to optimize ensemble fusion weights and decision thresholds for profit-driven prediction. The proposed framework could effectively improve ETP while maintaining competitive predictive performance. Compared with conventional prediction models optimized solely for classification accuracy, the proposed approach achieved a substantial improvement in customer retention profitability. The findings demonstrated that the optimized decision threshold significantly reduced the false negative rate in high-value customer identification. Furthermore, SHapley Additive exPlanations (SHAP)-based interpretation revealed that monthly charges, contract type, and tenure were the most influential factors affecting customer churn behavior. Profit-sensitive decision mechanisms provided a more effective strategy than traditional accuracy-oriented prediction approaches in customer churn management. The proposed framework provides a practical decision-support tool for intelligent customer retention and offers new insights into integrating business analytics with strategically operational decision making in the competitive telecommunications environments.