A Novel Framework for Mining Social Media Data Based on Text Mining, Topic Modeling, Random Forest, and DANP Methods

The huge volume of user-generated data on social media is the result aggregation users’ personal backgrounds, past experiences, and daily activities. This size generated data, so-called “big data,” has been studied investigated intensively during few years. In spite impression one may get from media, a great deal processing not uncovered by existing techniques engineering processing. However, very scholars have tried to do so, especially perspective multiple-criteria decision-making (MCDM). These MCDM methods can derive influence relationships weights associated with aspects criteria, which hardly be achieved traditional analytics statistical approaches. Therefore, in this paper, we aim propose an analytic framework mine networks, feed meaningful information via based theoretical framework, causal among finally compare theory. Latent Dirichlet allocation (LDA) will adopted topic models retrieved media. By clustering topics into theory, probability each aspect normalized then transformed Likert-type 5-point scale. Afterwards, for every topic, feature importance all other derived using random forest (RF) algorithm. matrix initial trial evaluation laboratory (DEMATEL). criteria DEMATEL-based network process (DANP). weight versus criterion DANP. To verify feasibility proposed Taiwanese attitudes toward air pollution analyzed value–belief–norm (VBN) theory Dcard (dcard.tw). Based results, are fully consistent VBN framework. Further, mutual influences work that were seldom discussed earlier works, i.e., between altruistic concerns egoistic concerns, as well those biosphere worth further investigation future.