A hybrid integration framework based on LOOCV and SARIMA: relationship exploring and predictive analysis between discipline attention and literature research

• Published in: PeerJ. Computer science Vol. 11; p. e2754
• Main Authors: Zhao, Yulin, Li, Junke, Liu, Kai, Shang, Chaowang
• Format: Journal Article
• Language: English
• Published: United States PeerJ. Ltd 01.04.2025; PeerJ Inc
• Subjects: Analysis; Big data; Computer Education; Data Science; Discipline attention; Forecasts and trends; Literature research; Multimedia; Network Science and Online Social Networks; Predictive analysis; Relationship exploration; SARIMA; Social networks; Visual Analytics; China; Discipline attention; Literature research; Predictive analysis; Relationship exploration; SARIMA
• ISSN: 2376-5992; 2376-5992
• DOI: 10.7717/peerj-cs.2754

Analyzing the relationship between the discipline of network attention and literature research can provide new insights for the innovative development of future disciplines. Many current studies focus on network attention, but its innovative application in the field of subject teaching has not been fully verified. Based on this, this paper proposed a relationship analysis and predictive analysis (RAPA) framework based on leave-one-out cross-validation (LOOCV) and Seasonal Auto-Regressive Integrated Moving Average (SARIMA) to explore the relationship between subject attention and literature research from the perspective of junior high school information technology. Based on the RAPA framework, five key keywords of this subject were extracted by combining the Baidu Index and China National Knowledge Infrastructure (CNKI) in first. Secondly, LOOCV was used to explore the relationship between subject attention represented by keywords and literature researches. Then, SARIMA was used to predict the future trends of subject attention and its literature researches. Finally, the prediction errors of different methods were compared. Based on the RAPA framework, the correlation analysis found that the r-values of subject attention and literature researches were all greater than 0.75, indicating a positive correlation between them. The predictive analysis found that the subject attention of junior high school information technology will be flat or decline in the next 2 years. Meanwhile, the amount of literature in this discipline has decreased compared to previous years, with an average of approximately 136. The prediction comparison showed that the prediction method in this study has a smaller mean absolute error (MAE) than other methods, and the MAE difference is 3.51. This indicated that subject attention, as an auxiliary variable of scientific research literature, is conducive to the quantitative analysis of literature research. At the same time, this study revealed the influence and role of big data represented by Internet attention in educational research.