管理工程学报2025，Vol.39Issue(5)：145-160,16.DOI:10.13587/j.cnki.jieem.2025.05.010
基于ELM理论和粒子群优化算法的在线评论有用性研究
Research on online review helpfulness based on ELM theory and particle swarm optimization algorithm
摘要
Abstract
The large number of online reviews accumulated on e-commerce platforms has become an indispensable information source for consumers and enterprises.Nevertheless,the exponential growth of reviews has led to information overload,significantly impeding consumers'ability to discern useful information.To address this,many platforms have implemented helpfulness-voting functions,enabling consumers to efficiently identify valuable information.In theoretical research,the prevalent method for quantifying review helpfulness is based on voting function,which has been widely accepted.However,this method presents certain challenges,such as the'cold start'problem for new reviews and the'Matthew effect'for older reviews.Additionally,this method overlooks the fuzzy nature of helpfulness,thereby risking inaccurate research results.Another mainstream method for quantifying helpfulness is based on text similarity,which effectively addresses the aforementioned problems.However,this method does not fully consider crucial information from the consumer's perspective,leading to insufficient satisfaction of consumers' needs for comprehensive and valuable information.This paper presents a novel method for quantifying review helpfulness based on the ELM theory and the particle swarm optimization algorithm,aimed at addressing the shortcomings of current online review helpfulness quantification methods.The method evaluates the helpfulness of review text in terms of the consumer's information-processing behavior pattern,which is more in line with consumers' actual perceptions.Moreover,the model effectively alleviates the problem of sparse helpfulness voting,thereby enhancing the quantification accuracy of review helpfulness. The first section succinctly surveys the pertinent literature concerning factors influencing helpfulness,various quantification methods,ELM theory,and particle swarm optimization algorithms.This review establishes a robust theoretical foundation for the research.The second section analyzes the characteristics of subjective and objective information within online review texts.According to the consumers' information processing patterns,it is summarized into central and peripheral path clues in the ELM theory to propose the research framework. The third section outlines the key steps of the experimental design.The first step is to collect the corpus and process the data.The second step uses the results of the LDA model to summarize six attributes of the hotel corpus and construct an attribute keyword dictionary to calculate the objective information content of online reviews.The third step calculates the subjective information content of online reviews based on the constructed sentiment dictionary.The final step proposes a quantitative model of online review helpfulness,introduces Kendall's harmony coefficient,and uses the particle swarm optimization algorithm as the basis to design the model parameter estimation and the calculation method for the review text helpfulness. The fourth section conducts three experiments using a dataset of 19,016 reviews from Qunar.com to validate the correctness,robustness,and effectiveness of the proposed helpfulness quantization and calculation methods.Specifically:1)Correctness test compares the parameters estimated using matlab and SPSS,which affirms that the model and method are correct.2)Robustness test uses various sample sizes to estimate parameters,which confirms the robustness of the results.3)Effectiveness comparison tests the effectiveness of the method by comparing the consistency of helpfulness based on voting,text similarity,and the proposed method with the real perceived review helpfulness based on the manual label.The experimental findings demonstrate the robustness and superior alignment of the helpfulness model and calculation method with the actual perceived helpfulness. Finally,the fifth section concludes the research and provides practical suggestions for online business platforms to enhance their online review functions.The proposed method for calculating online review helpfulness,based on the ELM theory and particle swarm optimization algorithm,effectively addresses the challenges associated with the cold start of new reviews,the Matthew effect of old reviews,and the sparse votes encountered in voting-based methods.Moreover,this method preserves the fuzzy nature of helpfulness and compensates for any deficiencies of text similarity-based methods,particularly for overlooking key information from the perspective of consumers.Furthermore,the findings reveal that consumers attach varying levels of importance to the perceived helpfulness of subjective and objective information when browsing online reviews.Objective information serves as the primary basis for their decision-making reference.Within subjective information,negative emotional content has a greater impact on perceived review helpfulness than positive emotional content.关键词
评论有用性/ELM理论/粒子群优化算法/主观信息/客观信息Key words
Online review helpfulness/Elaboration likelihood model/Particle swarm optimization algorithm/Subjective information/Objective information分类
管理科学引用本文复制引用
孙乔,吴锋..基于ELM理论和粒子群优化算法的在线评论有用性研究[J].管理工程学报,2025,39(5):145-160,16.基金项目
国家自然科学基金项目(71871177) The National Natural Science Foundation of China(71871177) （71871177）
