Answers are grounded in this paper and cite their sources.
Machine learning evaluation demonstrates accurate public opinion popularity forecasting in small-sample time series, indicating hybrid models effectively balance linear and nonlinear dynamics.
Online public opinion popularity exhibits strong temporal volatility, nonlinear behavior, and limited data availability, which pose challenges for accurate prediction using single models. To address these issues, this paper proposes a ridge regression - random forest fusion model for online public opinion popularity prediction under small-sample conditions. Using Baidu Index and Weibo interaction indicators, a composite popularity index is constructed via the entropy weight method. The proposed fusion model integrates the linear stability of ridge regression with the nonlinear modeling capability of random forest. Experimental evaluation is conducted using a real-world case study, with comparative analysis against ridge regression, random forest, BP neural network, and LSTM models. Results show that the proposed fusion model achieves superior prediction accuracy, with an R2 value of 0.9183 and a mean absolute percentage error of 6.99% on the test set. Five-fold cross-validation and multi-time-window robustness tests further confirm the stability and generalization ability of the model. The findings demonstrate that the proposed hybrid approach effectively balances linear trends and nonlinear fluctuations in small-sample public opinion time series, providing a practical and robust solution for online popularity prediction.
No takes yet. Share an insight, caveat, or question.
Guan et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: