基于微博公众情感状态的新产品市场预测研究

发布时间：2018-04-23 06:45

  本文选题：微博 + 公众情感状态　； 参考：《情报学报》2017年05期

【摘要】：使用社交网络数据获取公众情感,进而预测新产品市场趋势已成为社会化网络环境下市场信息预测研究的新方法。本文研究了基于中文微博情感分析的新产品市场预测的相关问题。首先,根据心理学的《心境状态量表》(POMS),从《同义词词林》中提取出七维度心境词汇种子词集;利用《同义词词林》及word2vec构建中文心境状态词汇语义网络,并通过马尔可夫随机游走算法计算词汇各心境状态维度值,自动化地构建了一个多维度、细粒度的情感状态词典,以便获取微博公众情感状态信息。进一步提出一个整合微博公众情感状态、微博提及数、评论情感及其数量的预测特征模型,采用多任务机器学习方法处理不同提前期的新产品市场预测问题。基于电影数据的实例分析表明,公众情感状态特征能在更长的时段内反映新产品市场趋势,且基于整合的预测特征模型和多任务机器学习方法具有较好的预测效力和预测提前期。
[Abstract]:It has become a new method to use social network data to obtain public emotion and predict the market trend of new products in the social network environment. This paper studies the market prediction of new products based on Chinese Weibo emotional analysis. First of all, according to the psychological state of mind scale (Poms), we extract the seven dimension mental state vocabulary seed words from synonym forest, and construct the semantic network of Chinese mood state vocabulary by using synonym forest and word2vec. A multi-dimensional, fine-grained emotional state dictionary is automatically constructed by using Markov random walk algorithm to calculate the dimension of mood state of vocabulary, so as to obtain the public emotional state information of Weibo. Furthermore, a forecasting feature model of integrating Weibo's public emotional state with the reference number of Weibo and commenting on emotion and its quantity is put forward, and the multi-task machine learning method is adopted to deal with the market forecasting problem of new products with different lead times. Case study based on film data shows that the characteristics of public emotional state can reflect the trend of new product market in a longer period of time. And the integrated predictive feature model and multitask machine learning method have better prediction effectiveness and prediction lead time.
【作者单位】： 华中师范大学青少年网络心理与行为教育部重点实验室;宾州大学神经成像中心;华中师范大学信息管理学院;孟菲斯大学智能系统研究所&心理学系;
【基金】：国家863计划基金项目“基于行为心理动力学模型的群体行为分析与事件态势感知技术”(2014AA015103) 国家自然科学基金项目“基于用户偏好感知的SaaS服务选择优化研究”(71271099),“基于屏幕视觉热区的网络用户偏好提取及交互式个性化推荐研究”(71571084)
【分类号】：F274;G206
，

本文编号：1790878