基于RF-BERT和UGC的用户需求识别及其发展趋势预测

User Demand Identification and Development Trend Prediction Based on RF-BERT and UGC

【目的/意义】通过社交网络平台中的用户生成内容(User-Generated Content,UGC)挖掘用户潜在需求,来帮助企业洞察市场变化趋势,进而优化研发产品和改进宣传策略。【方法/过程】首先利用TF-IDF技术和K-means算法提取在线评论中的产品属性特征,其次计算出每位用户评论中产品属性的情感值和关注度值后,通过RFBERT模型对用户评论中的产品属性特征进行筛选和分类。最后运用Bi-LSTM模型预测分类后的产品属性情感偏向和关注度的波动情况进而得到用户需求的发展趋势。【结果/结论】以“汽车之家”的在线评论为例,实验结果揭示出了用户评论被定义为精选评论时不同产品属性特征的影响程度,展示出影响值较高的属性特征的情感和关注度的波动情况。【创新/局限】论文提出了一种用户需求识别及其发展趋势预测的技术方案,为企业制定宣传策略和研发创新产品提供了参考。但选取的研究数据规模较小,后续研究中会扩大实验的样本注重研究结果的通用性。

【Purpose/significance】Mining user potential needs through user-generated content(UGC)in social networking platforms helps companies gain insight into market trends,optimize product development,and improve promotional strategies.【Method/process】Firstly,using TF-IDF technique and K-means algorithm to extract product attribute features from online comments,secondly,calculating the sentiment and attention values of each user's comments on the product attributes,and then filtering and categorizing the product attribute features in the user comments through the RF-BERT model.Finally,using the Bi-LSTM model to predict the fluctuation of product attribute sentiment and attention after classification,and thus obtain the development trend of user demand.【Result/conclusion】Taking the online comments of"qichezhijia"as an example,the experimental results reveal the influence degree of different product attribute features when user comments are defined as selected comments,showing the fluctuation of sentiment and attention degree of product attributes with higher impact values.【Innovation/limitation】The paper proposes a technical solution for user demand identification and development trend prediction,which provides a reference for companies to formulate promotion strategies and develop innovative products,but the scale of the research data is relatively small,and future research will expand the sample size and focus on the generality of the research results.