Information overload often makes it difficult for consumers to identify valuable online reviews through the traditional “helpful votes” button in the big data era, so it is essential to locate helpful reviews. Unlike the existing efforts that often measure online reviews’ helpfulness one-sidedly, this study takes the intrinsic helpfulness (IH) and extrinsic helpfulness (EH) into account, and the intrinsic-extrinsic comprehensive helpfulness (ICH-ECH) plot can be constructed by ensemble neural network model (ENNM) and time-weighted standard deviation accordingly. Furthermore, this study proposes a measure of EH ignored by previous studies, that is, the percentage of negative replies, which contain useful information that can measure online reviews helpfulness. We corrected it with a time sliding window by an improved iterative Bayesian probability approach (IBPA). In addition, this study further proposes a dynamic time-aware helpfulness ranking (DTAHR) model to dynamically rank reviews and identify beneficial reviews in a short time. We used real data sets from JD.com to conduct all experiments. The experimental results show that the performance of the DTAHR model is significantly better than other strategies. Our findings offer guidelines to evaluate the helpfulness of online reviews from multiple perspectives and rank them dynamically.

大数据时代，信息过载往往导致消费者难以通过传统的“有用投票”按钮识别有价值的在线评论，因此定位有用的评论至关重要。与现有的往往片面衡量在线评论的有用性的努力不同，本研究考虑了内在帮助性（IH）和外在帮助性（EH），内在-外在综合帮助性（ICH-ECH）图可以构建为集成神经网络模型 (ENNM) 和相应的时间加权标准差。此外，本研究提出了一个以前研究忽略的 EH 度量，即否定回复的百分比，其中包含可以衡量在线评论有用性的有用信息。我们通过改进的迭代贝叶斯概率方法 (IBPA) 使用时间滑动窗口对其进行了校正。此外，本研究进一步提出了一种动态的时间感知有用性排名（DTAHR）模型，用于在短时间内对评论进行动态排名并识别有益的评论。我们使用来自京东的真实数据集进行所有实验。实验结果表明，DTAHR模型的性能明显优于其他策略。我们的研究结果提供了从多个角度评估在线评论的有用性并对其进行动态排名的指南。实验结果表明，DTAHR模型的性能明显优于其他策略。我们的研究结果提供了从多个角度评估在线评论的有用性并对其进行动态排名的指南。实验结果表明，DTAHR模型的性能明显优于其他策略。我们的研究结果提供了从多个角度评估在线评论的有用性并对其进行动态排名的指南。