Interest points analysis for Internet Forum based on long-short windows similarity

• Main Authors: JU, Xinghai, LU, Jicang, LUO, Xiangyang, ZHOU, Gang, WANG, Shiyu, LI, Shunhang, YANG, Yang
• Format: text
• Language: English
• Published: Institutional Knowledge at Singapore Management University 2022
• Subjects: Point of interest (PoI) analysis; long and short windows; sequential analysis; deep learning; Numerical Analysis and Scientific Computing
• Online Access: https://ink.library.smu.edu.sg/sis_research/7185; https://ink.library.smu.edu.sg/context/sis_research/article/8188/viewcontent/TSP_CMC_26698_pvoa.pdf
• Institution: Singapore Management University

For Internet forum Points of Interest (PoI), existing analysis methods are usually lack of usability analysis under different conditions and ignore the long-term variation, which lead to blindness in method selection. To address this problem, this paper proposed a PoI variation prediction framework based on similarity analysis between long and short windows. Based on the framework, this paper presented 5 PoI analysis algorithms which can be categorized into 2 types, i.e., the traditional sequence analysis methods such as autoregressive integrated moving average model (ARIMA), support vector regressor (SVR), and the deep learning methods such as convolutional neural network (CNN), long-short term memory network (LSTM), Transformer (TRM). Specifically, this paper firstly divides observed data into long and short windows, and extracts key words as PoI of each window. Then, the PoI similarities between long and short windows are calculated for training and prediction. Finally, series of experiments is conducted based on real Internet forum datasets. The results show that, all the 5 algorithms could predict PoI variations well, which indicate effectiveness of the proposed framework. When the length of long window is small, traditional methods perform better, and SVR is the best. On the contrary, the deep learning methods show superiority, and LSTM performs best. The results could provide beneficial references for PoI variation analysis and prediction algorithms selection under different parameter configurations.