A major struggle of the museums is how to interest potential visitors and establish visitor loyalty with digital means. However, museum professionals offer limited insights into the actual visitor-visitor and visitor-museum interactions, which often happen beyond the museum space and are hard to observe. Nowadays, the interaction data between visitors and museums become obtainable because of social media. To conduct the analysis of visitor on SNS, we collected data from the official Instagram account of five most followed Korean museums and then, sampled a list of 5000 users from the followers of the museum. By visualizing the network and computing a number of network measures, we found the visitors can be clustered to six groups. For the behavior analysis, network analysis and Hashtag analysis were conducted to define the types and characteristics of the six groups. Based on this study, we will analyze actual visitor types and exhibition information element in real art museum.
Method and data description Top 5 Korean art museums having most followers on Instagram are selected as shown in Table 1: Daelim Museum; National Museum of Modern and Contemporary Art; Seoul Museum of Art; Ilmin Museum of Art; Kukje Gallery. 1000 users who followed the most recent are collected per each museum. To acquire user data from Instagram, urllib which is a Python package for working with URLs and Selenium WebDriver for automating web application are used. Since a private account is not available to identify user information, samples marked as private are removed from the dataset. To generate user network in terms of the connection between ‘following’ and ‘followed’, the following data, secondary followed account, of users in the dataset are gathered. Hashtags mentioned in user’s posts are also gathered to characterize users. A hashtag is a type of metadata used on social network services to make users find all the posts that have been tagged. Since users put hashtags to explain and indicate the content, the collected hashtags are regarded as a semantic feature. In the process of performing data mining, the combination of symbol characters is deleted from the hashtag list when it slices single word. Emoji characters, on the other hand, are preserved by encoding characters into UTF-8, e.g. 🎨, 📸, 🎬, 😍, 😁, ❤️, 💖, 🌊, ☕️, 🍕, etc.
Whole Network Analysis
– Statistical Analysis of whole social network
We have analyzed the social network of followers who follow five museums in South Korea on the Instagram as mentioned above. Figure 1. shows the relationship of 2838 followers. Originally it had a total of 5,000 nodes with 1,000 followers for each museum, but a total of 2,838 nodes remained, excluding private accounts and outliers. The outlier means that the user is not following or being followed. Thus, Table 2. shows that 2838 Nodes are connected to 1387 directed edges.
The statistical properties of our followers’ social network have been calculated in Fig.1.