In recent years, the incidence of public opinion in colleges and universities has been high. Monitoring, forecasting, and responding to public opinion of students in colleges and universities have increasingly become the work that education management departments at all levels attach great importance to. For each university, how to understand the sensation of teachers and students in real time in the era of informationization entering the intelligent campus has become an urgent problem. How to collect college campus network information, analyze and manage this information, and find hot topics from it has a profound impact on the reform of colleges and universities. Hence, in this paper, we propose a public opinion analysis framework based on intelligent data mining technique. Its advantage lies in the fact that it can withdraw the needed and unknown knowledge and regularities from the massive network data and host log data. It is a new attempt to use data mining in achieving public opinion. At present, data mining algorithm applied to public opinion analysis mainly has four basic patterns: association, sequence, classification, and clustering. Data mining technology is advanced for: it can process large amount of data. It does not need the users’ subjective evaluation and is more likely to discover the ignored and hidden information. Here, initially, the dataset is collected, which is preprocessed and divided into a training set and test set. Feature extraction of the text is done using Linear Discriminant Analysis (LDA). After that, text cosine similarity calculation is performed to compute the similarity between text vectors obtained from the LDA. Convolutional neural network (CNN) is used for classification purpose. We proposed Krill Herd Harmony Search Optimization Algorithm (KHHSOA) for optimizing the CNN and classifying the text into positive and negative opinion. The proposed system is simulated using MATLAB simulation tool, and the performance is analyzed in terms of metrics like accuracy, precision, recall, F -measure, kappa coefficient, and error rate. The proposed method is proved to be better when compared with the existing techniques.
Read full abstract