CPRP-CNN: A DEEP LEARNING FRAMEWORK FOR SHORT VIDEO POPULARITY PREDICTION USING IOT DATA

Abstract

Using deep learning regression models and the Internet of Things (IoT), this study looks at methods for predicting the popularity of short videos in order to provide useful insights and solutions that can be applied to content creators, social media platforms, academic research, and everyday users. In the context of cross-cultural communication, a proposed Convolutional Neural Network (CPRP-CNN) model-based Content Popularity Rank Prediction solely depends on the publisher's personal attributes and the textual features of short videos to predict the viewership levels of short videos shortly after they are released. The model's performance is evaluated by simulated tests, which show that using the Rectified Linear Unit (Relu) activation function in the CPRP-CNN model improves accuracy by 42.2% in comparison to using the sigmoid function. Alongside this improvement, cross-entropy loss is decreased by 37.8%. Additionally, compared to other prediction models, the suggested CPRP-CNN model achieves a cross-entropy of 0.692 and an accuracy of 74.7%, with better Mean Squared Error (MSE) and Mean Absolute Error (MAE) values of 2.728 and 1.751, respectively. These results indicate that combining deep learning models with fused features in an Internet of Things setting greatly improves the popularity prediction accuracy of short videos. The results of the study help to improve suggestions for accurate and tailored short video material.