ABSTRACT The reality about human behavior is that how other people think and evaluate have strong influences on our beliefs and thinking. Consumers get rich information from online reviews that may reduce their uncertainty regarding purchases. Besides, product-developing companies analyze user demands from online reviews to design market-driven product. In this study, a comparison among five major market share holder smart phone brands - Samsung, Apple, Huawei, Xiaomi, and Oppo is performed in different price categories - high, mid, and low range, based on sentiment polarity score. Online public reviews are extracted and sentiment scores of reviews are calculated to construct public sentiment polarity toward the famous brands. By examining both quantitative and qualitative methodologies, we identified the most important smart phone features or attributes that have great significance on consumer satisfaction. By experimenting and comparing five efficient machine-learning algorithms in predicting sentiment polarity and three feature selection algorithms in reducing attributes, an optimal set of 21 smart phone attributes was found those play major roles in determining customer satisfaction.

中文翻译：

---

识别产品特征对客户满意度的重要性识别公众情绪极性：使用机器学习方法分析智能手机行业

摘要 人类行为的现实是，其他人的想法和评价方式对我们的信念和思维有很大的影响。消费者从在线评论中获得丰富的信息，这可能会减少他们对购买的不确定性。此外，产品开发公司从在线评论分析用户需求，以设计市场驱动的产品。在这项研究中，基于情绪极性得分，对三星、苹果、华为、小米和 Oppo 五个主要市场份额持有者智能手机品牌在不同价格类别（高、中和低范围）中进行了比较。提取在线公众评论并计算评论的情感得分，构建对知名品牌的公众情感极性。通过检查定量和定性方法，我们确定了对消费者满意度具有重要意义的最重要的智能手机功能或属性。通过试验和比较预测情感极性的五种有效机器学习算法和减少属性的三种特征选择算法，发现了 21 种智能手机属性的最佳集合，这些属性在确定客户满意度方面起主要作用。