Retail promotion stacking and price confusion have intensified search frictions, exposing the fragility of conventional rule-based approaches. This study proposes a supermarket discount decision support system that replaces multi-agent architectures prone to intention drift with a dual-model collaborative framework, decoupling intent routing from conversational generation. The system integrates three key mechanisms. First, a structured interaction scaffold is constructed to reduce ambiguity in user inputs. Second, a dynamic fallback and replanning loop is designed to evaluate the confidence of CNN-based Optical Character Recognition (OCR) in real time, enabling autonomous global rerouting whenever confidence falls below a predefined threshold to improve robustness. Third, a memory table is introduced for data consistency verification, establishing a risk-control foundation through a traceable closed loop that spans image feature extraction, cross-validation against publicly available data, and confidence assessment. Experimental results demonstrate that, compared with the 8–10 operational steps typically required in manual workflows, the fully orchestrated system (Agent_full) consistently compresses highly constrained tasks into two interaction steps. The success rate of the T2 price comparison task reaches 86.96%, significantly outperforming B2 (46.15%) and A2 (73.08%). Although task completion time increases to 159.09 seconds, the system achieves higher-quality outcomes and minimal user interaction by shifting the cognitive burden of decision-making from users to system-level computation.

With the advancement of the digital era, community group buying has emerged as a new form of e-commerce. It has developed rapidly across China and has gradually become an important component of the retail market. As a major economic center in Northeast China, Dalian has witnessed the rapid growth of the community group buying market, which has not only transformed the consumption habits of urban residents but also demonstrated the vitality of this business model. The rise of community group buying has brought greater convenience to consumers while simultaneously creating both challenges and opportunities for traditional retail industries. However, with the rapid expansion of the market, community group buying has also encountered a series of problems, including insufficient transportation network coverage and unscientific route selection, all of which have affected its long-term development. This paper conducts an in-depth discussion of the current status of the community group buying distribution model in Dalian, the problems existing within it, and the corresponding countermeasures and recommendations. The study aims to provide valuable references for the development of community group buying in Dalian and across China, thereby promoting the healthy and sustainable growth of this model.

Against the backdrop of the current macroeconomic environment, financial institutions face mounting pressure to strengthen credit risk management throughout the lending process. This study introduces machine learning algorithms to establish a loan risk assessment system based on the Random Forest model. Based on the 'Credit Risk Dataset' published on the Kaggle platform, this study employed the Bootstrapping method to generate a subsample, and applied an ensemble of 100 trained decision trees to vote on the test dataset. Experimental results show that the model achieves an overall accuracy of 93.38% and an AUC as high as 0.9345, significantly outperforming logistic regression and single decision tree models. Furthermore, its high F1-score demonstrates that it has found a relatively ideal balance between "accurately identifying default applications" and "maximizing the interception of potential risks" which helps banks and other lending platforms maximize their profits while ensuring their own capital security.

The innovation patterns of cities in the Guangdong-Hong Kong-Macao Greater Bay Area (GBA) show prominent differentiation. Shenzhen takes original innovation and high-intensity Research and Development (R&D) investment as the core driver of urban innovation; Guangzhou focuses on scientific and educational resources and comprehensive innovation. Although Dongguan is a strong manufacturing city with R&D intensity ranking among the top in the GBA, it still faces shortcomings such as weak basic research, insufficient high-end talents, an imperfect achievement transformation chain, and inadequate collaborative innovation mechanisms compared with Guangzhou, Shenzhen, Hong Kong and Macao. By comparing the innovation capabilities of core cities in the GBA, this paper identifies the bottlenecks of urban innovation in Dongguan and proposes development paths, providing a reference for the construction of an innovative city in Dongguan.

E-commerce competition is shifting toward backend order fulfillment, and logistics capability has become a core driver of platform differentiation. Targeting the practical dilemma that self-built logistics incurs high costs in remote areas while Third-Party Logistics (TPL) boasts extensive last-mile coverage, this paper investigates two dominant modes of collaborative logistics: self-built logistics led mode and TPL led mode. E-commerce platforms are classified into resale platforms, marketplace platforms and hybrid platforms. By endogenizing users' perceived value of logistics services, this paper constructs game models and further introduces systemic risks to analyze the impacts of reliability losses on platform profits and mode selection. The results indicate that resale platforms prefer the self-built logistics led mode when users are highly sensitive to service quality; marketplace platforms tend to adopt the TPL led mode due to the lack of pricing power; hybrid platforms need to strike a balance between in-house service quality and third-party costs, and their optimal mode depends on the intensity of channel competition. When service sensitivity exceeds a certain threshold, the self-built logistics led mode generates higher profits. After incorporating systemic risks, a higher degree of aversion to reliability losses makes the self-built logistics led mode relatively superior on account of its stronger controllability. This research provides decision-making references for different types of e-commerce platforms to optimize logistics resource allocation.

In the current era of the integration of population aging and digitization, the scale of elderly Internet users continues to expand. New media has gradually become the core carrier for the elderly to obtain information, socialize and communicate, and consume and shop. However, new media marketing targeting the elderly still has many pain points: poor adaptability of marketing content, high operational threshold of platforms, uneven credibility of merchants, and difficulty in building trust among elderly users, which seriously restrict the high-quality development of the elderly digital consumption market. This article takes the elderly group as the research object, combines consumer behavior theory and precision marketing theory, and adopts the case analysis method to sort out the new media usage habits and consumption behavior characteristics of contemporary elderly people, analyze the existing problems and causes of new media marketing in the elderly market, and finally propose targeted optimization strategies from four dimensions: enterprises, platforms, regulation, and society. It aims to promote the standardized and humanized development of the elderly digital consumption market and achieve a positive integration of new media marketing and the elderly consumer group.

With the rapid development of the digital economy and content platforms, the influencer economy has gradually become a vital path for enterprises to expand markets, improve sales conversion and build brand influence. Based on the SWOT analysis method, this paper analyzes the mechanism of the influencer economy industry in boosting corporate profitability from four dimensions: strengths, weaknesses, opportunities and threats. The study finds that the influencer economy can improve corporate profitability through precision marketing, shortened consumption conversion paths, lower content marketing thresholds and enhanced consumer trust. However, enterprises also face problems such as over-reliance on a single influencer or platform, rising traffic costs, difficulty in turning short-term sales into long-term profits, weakened product competitiveness caused by excessive marketing, and intensifying homogenized competition. Furthermore, the improvement of digital infrastructure, the development of data technology, changes in consumer behavior and industry standardization provide opportunities for corporate profit growth, while tightening regulation, differences in social perception, changes in platform rules and macro consumption shifts also constitute external threats. On this basis, enterprises should give full play to the traffic advantages of the influencer economy, and at the same time strengthen product value, channel portfolio, data capability, compliance governance and long-term brand building.

Smartphone theft has surged alongside growing refurbished aftermarkets, yet standard pricing models overlook that thieves can observe product quality while buyers cannot. This paper analyses how a durable goods firm sets its price when buyers are uninformed about theft-protection quality, but thieves are fully informed, and a stolen product's resale value rises with the original price. This paper develops a three-player signalling game (firm, buyer, thief) and solves for equilibrium. The durable goods firm bears a per-incident operational cost if theft occurs. The study finds that when this cost is sufficiently high, a separating equilibrium emerges: the high-protection firm charges a premium that credibly signals quality, and theft is deterred. When the cost is low, a pooling equilibrium prevails: both quality types set the same price, quality remains hidden, and low-protection products are stolen. Applying the Intuitive Criterion refines the separating outcome to a unique prediction. The model shows that a firm's own operational theft cost is the critical determinant of whether price can serve as a reliable quality signal, with implications for firm strategy and theft-liability policy.

This paper empirically explores the impact of Environmental, Social, and Governance (ESG) incidents on stock returns. Utilizing a dataset spanning 16 years from RepRisk, this research analyzes the daily abnormal returns of 24,547 listed companies worldwide. The results indicate that the market reaction to ESG incidents is generally short-lived, with significant price corrections occurring primarily within a three-day window following the event. This research documents a distinct heterogeneity across ESG pillars: Governance (G), specifically those involving anticompetitive behavior and controversial products and services, precipitating immediate and significant stock price declines. Conversely, Environmental (E) incidents show statistically insignificant short-term impacts. While the aggregate Social (S) category shows mixed results, specific sub-categories such as forced labor and animal cruelty trigger significant negative market reactions. In summary, while ESG receives widespread attention, short-term market volatility is predominantly driven by governance failures.

To systematically investigate the mechanisms and realization pathways through which New Quality Productive Forces (NQPF) facilitate the development of Gansu's modern industrial system, this study employs balanced panel data for 12 prefecture-level cities in Gansu Province from 2012 to 2024. A hierarchical entropy-weighting method is adopted to quantify the levels of New Quality Productive forces (NQP) and Modern Industrial System (MIS) development. Two-way fixed-effects, mediation-effects, and interaction-effects models are subsequently constructed for empirical testing. The empirical results indicate that new quality productive forces significantly promote the development of Gansu's modern industrial system. Both digital technological innovation and green technological innovation serve as mediating mechanisms, with the latter exhibiting a more robust transmission effect. Moreover, digital and green technological innovations demonstrate a significant positive synergistic effect. Heterogeneity analysis reveals that the empowering effect is concentrated primarily in cities with higher levels of economic development, non-resource-based cities, and cities with a stronger secondary industrial base. In addition, a central-city spillover pattern is observed, while the release of policy effects exhibits a certain time lag.