Construction and Data Analysis of a New Media Content Popularity Prediction Model Based on Naive Bayes Algorithm

In view of the difficulty of predicting the popularity of content on new media platforms, the Naive Bayes algorithm is introduced to accurately predict the popularity and dissemination trend of content by analyzing user behavior and content characteristics. This paper constructs a popularity prediction model based on Naive Bayes and verifies it through actual data, demonstrating the effectiveness of the model in new media content analysis. First, the historical content data on a new media platform is collected and cleaned, including feature information such as article titles, tags, publishing time, user interactions (likes, comments, shares, etc.). Secondly, the Naive Bayes classification algorithm is used for modeling, and the text features of the content and user interaction data are used as input variables to predict the popularity of the content. The SMOTE (Synthetic Minority Over-sampling Technique) method is used to oversample the training data. By training the model, this paper obtains the probabilistic relationship between various features and popularity. Finally, the cross-validation method is used to evaluate the accuracy of the model, and the prediction results are adjusted and optimized according to the actual data. The short-term, medium-term and long-term features under text + user behavior features + time series features (long-term) respectively improve the model performance to 81.3%, 82%, 82.5% accuracy, F1 value to 0.73, 0.74, 0.75, and recall rate to 79%, 79.5%, 80%. By constructing a popularity prediction model of the naive Bayes algorithm, this paper proposes a new data-driven analysis method, which can provide theoretical support and practical guidance for content recommendation and marketing strategies of new media platforms.