Short Video E-Commerce Marketing of Convolutional Neural Network Based Transfer Learning with Visual Geometry Group16

In Recent years, E-commerce stands as a prominent testament to the transformative power of information and communications technologies in driving economic growth and development. Traditional approaches from short video ecommerce marketing had faced several issues which include large data size, required more computational resources. Therefore, this research proposes Convolutional Neural Network (CNN) based Transfer Learning with Visual Geometry Group (VGG)16 for Short Video E-commerce Marketing. Initially, the data is taken from Buyer-generated Fashion Video (BFV) dataset and consists of fashion videos generated by users, showcasing their purchases and experiences with fashion products. The data is preprocessed by using frame extraction and then the features are extracted by using CNN based Transfer Learning with VGG16 which effectively reduces the training time and computational resources. Next, the You Only Look Once (YOLO) is used for feature engineering where YOLO enabled the analysis of object based features. Finally, the evaluation and optimization is done by Multi armed bandit where it provides efficient way to balance exploration and exploitation which enabled real time optimization. The proposed CNN based Transfer Learning with VGG16 achieved accuracy (0.955), precision (0.977), recall (0.968) and F1 Score (0.970) when compared with existing Hierarchical Oversampling Word Embedding Network (HOWNET).