2020年美国大学生数学竞赛优秀论文：情感分析与ARIMA模型在市场趋势预测中的应用

版权申诉

92 浏览量更新于2024-06-15 收藏 2.54MB PDF 举报

"2020年美赛优秀论文集-C-2009116.pdf"主要收录了美国大学生数学竞赛中的一项获奖作品，其关注的核心是数学模型在实际商业问题中的应用，特别是市场数据分析。论文详细探讨了如何利用Amazon的客户评分系统作为案例，通过情感分析（Sentiment Analysis）模型和ARIMA模型来挖掘和预测产品市场趋势。研究者针对原始数据中的缺失值、异常值以及重复观测进行了数据预处理，采用了定性与定量方法相结合的筛选策略，剔除无关变量。论文首先运用描述性统计来初步了解产品的市场满意度状况，这有助于构建一个清晰的产品市场画像。接着，作者采用基于词典的情感分析技术，对客户评论进行深度剖析，构建了一套情绪评价体系。这个系统能够快速捕捉并量化消费者对产品的正面或负面情感，为公司决策提供依据。通过这种结合了定量分析（如ARIMA模型）和定性分析（如情感分析）的方法，研究者实现了对市场动态的深入洞察，从而帮助企业更好地把握未来发展趋势，制定具有竞争力的商业策略。这篇论文不仅展示了数学模型在解决实际商业问题中的威力，还强调了数据处理和分析在现代市场竞争中的关键作用。对于那些对大学生数学竞赛、数学建模以及市场数据分析感兴趣的人来说，这是一份极具参考价值的资料，对于提升商业智能和决策支持能力具有重要的指导意义。"

关注数学模型

获取更多资讯

MATHmodels

Team # 2009116 Page 5 of 24

1 Introduction

1.1 Statement of the problem

Online marketplace where Amazon takes the leading place has become the most

important platform to trade for most companies. As the mechanism for customers to

communicate shopping experience with each other, various kinds of reviews and ratings

provide potential consumers with subjective evidence. It can also provide the guidance

for new entrants in choosing a target market and target product. In addition, in the social

media era, consumers have louder and stronger voices than ever before, which is shown

as over 50% of consumers frequently factor in online reviews before buying a

product.[1] Because this process can be seen as a direct link to the former customer,

getting the first-hand information of product in order to avoid the exaggeration or of

the seller.

However, at this stage, it is not clear that their interaction, and how they influence

the corporate strategy. Thus, we aim to explore the internal relationship of three

variables and design a composite data measure to assist companies in the selection of

potential successful products. Meanwhile, we discuss how the customer group reviews

influence on individual review attitude orientation. At last, we rank the alternative

products according to our method and offer a recommendation list to Sunshine

company.

1.2 Literature Review

At present, there are mainly two methods for sentiment analysis, one is based on

machine learning, and the other is based on the lexicon. Several approaches based on

the machine learning have come into the spotlight in recent years and most of them are

based on the big amount of personalized reviews from social media. Due to the reviews

in the social media are not only English, but other languages such as Arabic. [2] Many

researchers from non-English regions also apply this method to solve the issue, showing

that this method is quite mature.

And the other method is the Lexicon based approach. This method also utilizes the

text which expresses people’s sentiment or emotion in their social media, such as the

Twitter posts and Micro-blog text. [3,4] Latest researches on the sentiment polarity

lexicon also show the resource of the lexicon is extended to a particular domain, the

stock market, without human intervention and addressing the scaling and thresholding

problem. [5] Therefore, we believe that this method is potential to be widely used.

After we compared the two methods, we determine to choose the Lexicon-based

approach. We believe that this method is most applicable when time and resources are

limited, that is, similar to the conclusions obtained in the study of Urdu's Sentiment

Analysis. [6]

After introducing the time factor, we propose a time series model to discuss the

changes of three variables in the time dimension. After comparing several applicable

models, we choose to use the traditional time series models and the Autoregressive

Integrated Moving Average (ARIMA) model to observe the influence of time factors

on different reviews and to predict the future indicator trend. We think that during the

forecasting phase, the ARIMA model can achieve a quite accurate forecasting as

theories of Matyjaszek, M. et al. [7] At the same time, we also conducted product

forecasting of different products by product. This is in line with Nguyen, H. et al. The

model's predictions for different product prices are similar [8].

1.3 Overview of Our Work

Our main goals are to address three issues that need to be explored: (1) the

sentiment analysis of text reviews; (2) the relationship between reviews and ratings

indicators and (3) the impact of time. Then we propose the product sequencing method

which these three contents combined with.

In order to solve these problems, we first analyze the data characteristics of each

digital parameter, and then apply the sentiment lexicon-based method to analyze the

emotional attitude of the text of the text review, extract the emotional words in the

剩余23页未读，继续阅读

阿拉伯梳子

粉丝: 2185
资源: 5737

2020年美国大学生数学竞赛优秀论文：情感分析与ARIMA模型在市场趋势预测中的应用

2020年美赛优秀论文集

2020年美赛优秀论文.rar

2020美赛C题O奖2002116论文（中文版）.pdf

全球气候变化政策建模：2020年美赛优秀论文

2020年美赛论文合集：生态与渔业经济的迁徙预测模型-研究苏格兰鲱鱼与鲭鱼的影响

java基于ssm+jsp一家运动鞋店的产品推广网站系统源码 带毕业论文

51单片机Proteus仿真LCD1602+DS18B20的温度读取显示编程.rar

暴风电视 50F1 配屏V500HJ1-PE8(C3) 机编600000MWV00 屏参30162503 风UI1.0 本地升级

【高创新】基于蚁狮优化算法ALO-Transformer-BiLSTM实现故障识别Matlab实现.rar

java基于ssm+vue儿童影楼拍摄管理系统源码 带毕业论文

最新资源

java基于ssm+jsp一家运动鞋店的产品推广网站系统源码带毕业论文

java基于ssm+vue儿童影楼拍摄管理系统源码带毕业论文