Under Korean Eyes: Expat Vlogging on Thailand

Korean YouTuber Cullen and friends have been the talk of the town after producing travel videos that foreground the host-guest relationship, according to data analysis.

Hailing with friends from Korea, Park Kideuk, known as Cullen, is a music producer and member of the trio band "HateBerry", which also features Kirby and Pino. He uploaded the first video on his YouTube channel “Cullen HateBerry” on April 7, 2022 to show his everyday life in Thailand.

As of the data collection date on Jan 18, 2024, there were 93 videos and more than 1.7 million subscribers.

My capstone project is aimed at finding out why the channel is going viral. According to Google Trends, public interest based on the search term has dramatically increased since Nov 2023. Top related queries include his brother Jung.

Finding Patterns in Data

After scraping data from his channel, I spent many days cleaning and organising them on three spreadsheets: video, playlist, and comment.

My analysis identifies the correlation between views and two other metrics. The first scatter plot shows that the number of views corresponds to that of episodes. Its curved trendline surpasses 1 million views in the second half of all videos.

The second scatter plot demonstrates that the number of views is proportional to that of likes, as shown by its straight trendline.

The last scatter plot points out that the number of views increases with the duration of each video.

It should be noted that an outlier is shown in the top right-hand corner of each scatter plot. As of the data collection date, the most popular video titled “Judy! Let’s have a fun trip to Suphan Buri (Ft.Loy Krathong)” (จูดี้!พาทุกคนไปทริป แห่งความสนุก! ณ สุพรรณบุรี (Ft.ลอยกระทง)) rakes more than 5.6 million views.

Provincial Trip

According to the chart, travel seems to be the most popular content. To prove my hypothesis, I divided videos into two groups, with the result that the average view of each provincial trip, some of which contain several episodes, is higher than that of urban lifestyle.

By classifying provincial trip into regions, I found that the average view of the northeastern region peaks at 2.6 million.

Host-Guest Relationship

At first, I requested sentiment analysis of over 30,000 rows of comments from all videos, but ChatGPT 3.5 was unable to cope with an avalanche of the data provided. I scaled it down to 500 comments from the most-viewed video. Fortunately, ChatGPT could live up to the smaller task. It shows their overwhelmingly positive response and the most recurring words.

Contrary to popular belief, Judy is reported with higher frequency than Cullen and Jong, according to the sampled group of comments. Then I read them to categorise contexts in which these words appear. People speak highly of Judy’s role in introducing foreigners to provinces and their customs. Meanwhile, Cullen and Jung are noted for their curiosity and gentleness.

In addition to travel content, the ideal host-guest relationship arguably makes the program popular. Judy acts as a tour guide who helps connect foreigners to villages. For example, she led them to pick ingredients from the farm. Meanwhile, Cullen and friends are examples of tourists who are humble and open to local culture, such as learning to speak Thai.

Feedbacks

During my pitching, an instructor questioned whether the surge of views necessarily results from the number of episodes. It gives an impression that the more videos creators churn out, the more popular they become, which is not true. In his own analysis, the growing number of views correspond to that of subscribers.

He also doubted the accuracy of data collection, saying that an upgraded (paid) version of ChatGPT can handle a larger volume of data. Better prompts can also improve sentiment analysis.

He said the last part is quite qualitative, but my attempt at sentiment analysis is interesting.

P.S. This summary is part of my capstone project during the 7th Data Analytics Bootcamp at Skooldio from Nov 2023-Feb 2024.