The digital transformation has revolutionized the marketing landscape, offering both opportunities and challenges to marketers and consumers alike. In India, the retail industry is projected to witness substantial growth, with online retail expected to reach billions by 2030. This growth is driven by the rise of online social commerce, fueled by the increasing number of users on social networking sites such as Facebook, Instagram, Pinterest, and Twitter.

This research aims to explore the changing perception and role of social commerce among two generational cohorts, Millennials and Generation Z, in the Indian market. The study employs descriptive research methodology with a structured questionnaire administered to 200 respondents through purposive sampling. The findings reveal that social media has become an integral part of consumers' lives, with Instagram and YouTube being the most preferred platforms among both generational cohorts. Consumers spend considerable time on social media, presenting a promising opportunity for marketers to engage with their target audience.

Social media influencers play a significant role in influencing consumer behavior, particularly among Generation Z, who show higher trust in the information provided by influencers. Millennials tend to rely on influencers for specific product/service categories, while Generation Z is more open to influencer content across various domains. Despite their positive attitude towards influencer content, there is a discrepancy between consumer attitude and actual purchase behavior. While Millennials are more likely to consider market offerings endorsed by influencers, Generation Z uses influencer content primarily for information gathering and relies on other factors for final purchase decisions.

The managerial implications of this research highlight the need for marketers to collaborate with relevant influencers, focus on delivering high-quality content, and leverage multiple social media platforms to engage with consumers effectively. Marketers should also address the gap between consumer attitude and behavior by tailoring influencer strategies for different product segments and monitoring campaign performance. By understanding the specific role of social media influencers and designing effective marketing strategies, businesses can capitalize on the opportunities presented by social commerce in the dynamic Indian market.

Aiming at the existing deep learning-based flame detection methods, which have problems such as slow detection speed, large model difficult to deploy to edge computing devices and poor flame detection effect for small targets, an improved model based on YOLOv5 is proposed and applied to flame detection. Firstly, to address the problem of the model being large and difficult to deploy, this paper proposes a lightweight evolution method based on GhostNet; in order to further improve the accuracy and the effect of detecting small target flames, Coordinate Attention is added to the backbone network, which makes the network pay more attention to important information and reduces the influence of useless information, thus improving the performance of the network; finally, the EIoU Loss instead of CIoU Loss as the loss function of the algorithm, which improves the localization accuracy while increasing the rate of bounding box regression. The experimental results show that the new model obtained based on the above improvement method has obvious advantages, the model size is only 7.6MB, which is 45.3% smaller than the original model; the inference speed reaches 99.4FPS on RTX3080 device, which is 12.8% higher than the original model; and the mAP reaches 0. 881, which is 2.3% higher than the original model

The rapid development of the Internet has brought great convenience to people's daily life. However, the fraudulent behavior of lawless elements through phishing links has become more and more intense, and has seriously jeopardized the safety of people's lives and property. At present, the main technologies for detecting phishing links are based on blacklists, machine learning, but these technologies require a lot of manual labeling, which is time-consuming and unstable. After an in-depth study of the phishing link problem, we propose a network model TTCN, which firstly embeds URL links at character level and word level, then extracts feature representations from TCN and Transformer respectively, fuses these feature representations, and classifies them through fully connected output. The experimental results show that this model achieves 93.61% accuracy in recognizing phishing links, which can cope with the fraud problem generated by phishing links and helps to maintain network security and protect people's lives and properties.