we handed out questionnaires to 227 students from 8 classes of 3 rural primary schools. In this questionnaire, the students were asked to list 5 classmates they like most and 5 classmates they dislike most. We use the information to constructed their social network as a graph and visualized it to get insights.

The raw data is not provided here for privacy reasons.

The network of "like" and "dislike" relationships were construted as a directed graph using . The visualization was completed by .

The social network graph of all the students, in which each color indicates a class.

The "like" relationship was indicated by red arrows and "dislike" by green arrows. Girls are represented by red nodes while boys by blue.

By running the commutiy detection algorithm, we found that the social network within a class was splited by gender. This is reasonable considering the age of the students and the culture of Chinese rural areas.

This kind of students are the focus of the whole class. Usually, they are "liked" by one community while "disliked" by another.

This kind of students have more mutual "like" relationships than others.

They are not liked or disliked by anyone.

• Quantative analysis of the relationship of the background of the students and the social status of the students in school.
• Develop alarming systems to identify students who might need special attention by analysing social network.