If you have been following up to this point, you may get very poor results from all the clustering algorithms. I'd like to remind you that the stated objective of this book in general is to impart an understanding of what it's like to do data science in Go. For the most part, I have advocated a method that can be described as think hard about the problem, then write the answers down. But the reality is that often trial and error are required.

The solution that works for me on my Twitter home timeline may not work for you. For example, this code works well on a friend's Twitter feed. Why is this? He follows a lot of similar people who talk about similar things at the same time. It's a little harder to cluster tweets in my Twitter home feed. I follow a diverse array of people. The people I follow don't have set schedules of tweeting and do not generally interact with other Twitter users. Therefore, the tweets are generally quite diverse already.

It is with this in mind that I encourage you to experiment and tweak your program. In the subsections that follow, I shall outline what worked for me. It may not work for you.