In the medicinal field, predictive models are critical for understanding and trust. Various current machine learning (ML) methods, such as those employed for the automated prediction of depression, are difficult to explain. In this sense, this paper proposes a social media (SM) user behavior prediction (UBP) method employing the Variational Onsager Neural Network (VONN) model to challenge the dynamics of user behavior patterns, and a Mud Ring Algorithm (MRA) to improve the features of VONN. The proposed method uses demographics, data on social interactions, and content choices to effectively anticipate user behavior. The analyzed data was first gathered on a large scale using trustworthy ground truth datasets, and then it was preprocessed utilizing the Switched Mode Fuzzy Median Filter (SMFMF) approach for refining the data on user behavior. The Binary Ebola Optimization Search Algorithm (BEOSA) was also used for optimizing the characteristics of the social media textual content and users’ posting behaviors. The improved data was then input into the MATLAB-implemented VONN model, which classified users as depressed or not depressed. Performance indicators included accuracy, the F1-score, sensitivity, specificity, precision, recall and ROC, and were used to investigate the proposed technique’s performance. This paper extended the technical landscape related to social media user behavior prediction by addressing the key problem of model explainability in the context of automated depression prediction. The proposed UBP-SM-VONN-MRA achieved an accuracy of 98.6% in detecting depressed users and 99.8% for non-depressed users, which demonstrates its commitment to technological excellence in the quest for precise and understandable predictions. The other three employed techniques, namely UBP-SM-RNN, UBP-SM-ANN, and UBP-SM-MDHAN obtained a lower accuracy. The obtained simulation results showed that the UBP-SM-VONN-MRA technique outperformed the other three techniques and that it has a greater potential for identifying depression.
Binary Ebola Optimization Search Algorithm (BEOSA), Depression Detection, Social Network, Mud Ring Algorithm, Variational Onsager Neural Network, Switched Mode Fuzzy Median Filter (SMFMF).
Xiaomei ZHANG, Peipei QI, Can XU, Dandan SUN, "Social Media User Behavior Prediction Based on VONN and MRA", Studies in Informatics and Control, ISSN 1220-1766, vol. 33(4), pp. 113-123, 2024. https://doi.org/10.24846/v33i4y202411
All articles published by Studies in Informatics and Control (SIC) are published in accordance with the terms and conditions of the Creative Commons Attribution License (CC BY-NC-SA 4.0).