تحلیل نظرات کاربران وب سایت های تجارت اجتماعی بر اساس روش های متن کاوی و داده کاوی

نویسندگان

1 استاد تمام دانشکده مدیریت دانشگاه تهران

2 استادیار دانشکده مدیریت دانشگاه علامه طباطبائی

3 دانشجوی کارشناسی ارشد دانشکده مدیریت دانشگاه تهران

چکیده

هدف از این پژوهش مطالعه و تحلیل نظرات کاربران وب­سایت­های تجارت اجتماعی با بهره­ گیری ترکیبی از تکنیک ­های متن­ کاوی و داده­کاوی است. بدین منظور پایگاه داده وب سایت تریپ ادوایزر مورد ارزیابی قرار گرفته و از میان همه کاربرانی که در مورد یکی از هتل­های شهر لیسبون پرتغال ابراز نظر کرده­ اند، نمونه ­گیری سیستماتیک انجام می­شود. سپس همه نظرات این کاربران نمونه در مورد همه مطالب از قبیل هتل­ها، رستوران­ها و اماکن دیدنی استخراج شده و با استفاده از نرم ­افزار رپیدماینر و تکنیک­ های متن ­کاوی و داده­ کاوی تجزیه و تحلیل می­ شوند. محاسبه شاخص دیویس و بولدین بهینه ­ترین تعداد خوشه، سه به دست می­آید و کاربران به سه خوشه تقسیم می­شوند. هر خوشه دارای ویژگی­ های منحصر به فردی است که پرتغال ­دوست، رستوران ­دوست و سفر­دوست نامیده می­شوند. سپس بر اساس ویژگی­ های هر خوشه و اطلاعات موجود در پروفایل کاربران با استفاده از قواعد انجمنی اقدام به شناخت بیشتر هر خوشه می­شود. در انتها متناسب با ویژگی­های تحلیل شده هر خوشه اقدام به ارائه راهکار جهت افزایش مشارکت آن­ها در وب سایت کرده، و پیشنهاد­های ترفیعی هدفمندی برای هر کدام از این خوشه ­ها بیان می­شوند.

کلیدواژه‌ها


عنوان مقاله [English]

An Analytical Approach to Social Commerce Websites’ Consumer Reviews Based on Text and Data Mining Techniques

نویسندگان [English]

  • Babak Sohrabi 1
  • Iman Raeesi Vanani 2
  • fereshteh Khodaparast 3
1
2
3
چکیده [English]

Received: 29/12/2016

Accepted: 17/03/2017

Extended Abstract:

Introduction and statement of the problem: Nowadays customers do not visit online shopping websites alone, but they bring all of the social networks capabilities with themselves. This cooperative environment made users as active content providers despite being passive readers in web. Customers in this area not only purchase services or products but also they provide contents simultaneously and it will cause a win-win relation for sellers and customers.

Customers do not have opportunity for smelling and touching products in online shopping, so the comments of the others are getting more important every day. Especially when the person writes the comment, he or she uses or touches the product or service closely. Indeed, these comments and rankings are a vital part for potential customers. According to such issues, the main problem of this study is to investigate the users’ behavior in social commerce websites, focused on tourism industry. Based on the content provided and the performance analysis of users, some approaches can be presented for increasing the users' cooperations in social commerce websites, using data mining and text mining techniques. Finally, media and tourism industry can use and enjoy the results of this study.

Theoretical background: Some studies show that potential customers are more interested to review comments and proposals from other people than information provided by producers or service providers. A significant part of economic activists started entering this section of e-commerce regarding huge business opportunity occurred by e-commerce. Market experts start using this media to maximize their profits.

For the first time, Yahoo used social commerce term for describing its cooperating online shopping tools and ranking by users. Yahoo decided to make a society of buyers who rank products, share their experiences, and finally present this information to end-users.

Social commerce does not have an exact meaning because it has different conceptual utilization for different people. Generally, it has been used as a subset of e-commerce, which supports social media for online selling and purchasing products or services.

Social commerce might be applicable for business to consumer, business-to-business, and consumer-to-consumer relationships. The role of media in tourism and industry is very complicated and important as statistics show that most tourists in the world are people who visit a destination for the first time. They gather all required information about the trips from various media and this is the reason why media are getting more and more important in tourism stream. Therefore, it is necessary to plan to use this opportunity.

Methodology: This research used library, papers, and books study for gathering information from broad literature reviews. We did systematic sampling from among all users and took comments about all hotels in Lisbon city in Portugal at an accredited online tourism-centric website. The number of initially selected participants is 2681 and the number of final accessible sample is 535 users. After extracting comments from the website, we analyzed the data using data mining and text-mining tools by rapid miner software and then users have been deeply clustered and analyzed. CRISP-DM is the main method that is used in the research containing 6 steps: business understanding, data understanding, data preparation, modeling, evaluation and deployment.

Results and discussion: Davies-Bouldin index is used for validation and determining the optimized number of clusters. The value of the index for three clusters was about 0/003 and three clusters were selected as the optimized number of clusters. These clusters are finally named as Portugal lovers, restaurant lovers and trip lovers. Users in Portugal lovers cluster focused specifically on Lisbon city in Portugal. Most of these users were from Portugal or that country was one of a few destinations for their main trips. They are very sensitive on location of hotels and restaurants and it is one of their major priorities and concerns that the places are in city centers. On the other hand, restaurant lovers cluster are very interested in restaurants and their locations. They provided comments specifically on food, quality of service and the behavior of staff. For these people, friendly behavior of staff was very important. The people were more romantic than other clusters and maybe for this reason, they spent most of their time in various restaurants and cafes. In trip-lovers cluster, the focus was on hotels and accommodation. These users prefer to provide comments on hotels, their specifications, and the cleanliness of equipment. Besides, this cluster contains most of the couples.

Based on association rules analysis, Portugal-lovers cluster travel only to few destinations and they are inactive readers in the website and do not cooperate for commenting on their personal experiences. Restaurant-lovers people are the best in cooperating to share their experiences in the website. For these users, the main purpose of traveling is food and restaurants. People in trip-lovers cluster are in average levels of cooperation and they provide comments on hotels more than other two clusters.

Conclusion: For Portugal lovers cluster, researchers should focus on Portugal hotels and restaurants proposals. They should propose packages for composed hotels and restaurants in Portugal because they have rather the same priority for them. For more contribution to the websites comments, they can be provided with some bonuses to be used in the favorite places in Portugal while using the website.

Because restaurant lovers cluster has the best contribution levels in the website, hotels should plan an extra promotion to retain them based on their favorite locations. For more contribution for trip-lovers cluster in the website, researchers and practitioners should propose clean hotels with friendly staff and good breakfast, because they are professional and permanent travelers.

Prominent results: Users contributing in social media have two important dimensions to study. The first is the profit of social media and the second is increasing contribution level of users in such media for increasing the effective impact of sharing. These two aspects are positive loops that amplify each other. In other words, more users' contribution in social media, that makes accurately extracted analytical results, can cause more validity for that media. Therefore, if social media receive valid and useful contributions, number of permanent social users will be increased and this trend will continue to produce fruitful information to decide upon.

کلیدواژه‌ها [English]

  • Social commerce
  • E-commerce
  • Social media
  • Data Mining
  • Text mining