Not exactly. I did some tests, and initially I wanted the search to be performed in a "network" starting from the user who is doing the search, considering the "distance" within that network as a factor for relevance — meaning, in a search, users closer to me, with more similarity to my query, would have higher relevance, and those farther away would have lower relevance.

In theory, it looks great, but when I tested it in practice, perhaps because the number of users was still small, the results were quite poor. So, I added a "similarity" property to the users resulting from the search and simply sorted the list at the end in descending order by the "similarity" property.

In the form described above, the results were much more accurate, so once I got the desired outcome, I optimized the code to make the search as fast as possible. To summarize, the most efficient approach given the current number of users is to take the list and perform a simple search sorted by similarity. I’m still not making good use of the graph properties, and I’ll revisit the analysis when the user base is large enough for graph-based search results to be more relevant.

This week, I’ll solve an issue in the method that calculates the similarity between users’ names in the list and the search term. There’s a bug causing it to return poor results quite often.