Pōsōh
I hope everyone has been well over the last 4 months! I'm thrilled to share that my research commenced in May and has progressed significantly in the last two months. My research is titled "Community Detection of Indigenous Beaders on Instagram," and I'm proud to say that my understanding and confidence in my topic and my research has exponentially grown in the last month. I can clearly define community detection, explain how decoding works, and navigate Python far better than I could have imagined possible at the end of April. Today, I will share my research process and further questions I have about my research.
Research Process
To begin my research, I needed to gather my data. I collected 50 different accounts of Indigenous people who sell beadwork on Instagram. Then, I saved each of their usernames into a list in the same folder as my Python programming page. Next, I proceeded to save the list of followers each of the accounts had, and once again stored it in the same folder. Later, I created an algorithm in Python that attributed each list of followers to the correct account and proceeded to ensure that Python erased all the followers that were not affiliated with the specified list of 50 accounts. This allowed me to write the most important line in my code, the double follow code. The double follow code listed each pair of users that mutually followed each other on Instagram and assigned an edge between them on my graph. Using this information, I was able to create my adjacency matrix! The adjacency matrix is the key piece of data we need to begin using our Maximum Likelihood Decoding Algorithm. The graphs pictured below are images of the data we collected. Each node represents an Instagram account, and each edge represents a double follow. The adjacency matrix will show a 1 in each place where an account has a double follow, and a 0 where there is not. We will begin using Maximum Likelihood decoding techniques this week to find whether there are smaller niche communities within our larger community. However, at this time we have not decided which graph is able to help us better interpret our data.
Further Questions
One thing I love about doing research is that I get to decide the format of my work and how my graphs should be presented. Unfortunately, the decision-making process can be difficult because whatever format I choose will lay the foundation for how others create their graphs in the future. Hence, I aim to choose a format that is easy to recreate and easy to interpret. Unfortunately, at this time, my choice of graph is difficult to read and difficult to recreate. My question is: How can I find a graph formation that naturally separates into clusters that are more connected? Additionally, before I start using maximum likelihood decoding techniques, I want to know what an ideal size a community should be and how many communities I should expect to form from a group of 50. Is one person too small for a community? Using maximum likelihood decoding, I get to choose these parameters and decide the minimums and maximums. I'm hoping to have answers to these questions in the next few weeks. For now, those are all of my updates.
Wāewāenan
Comments