Though the conclusion that this algorithm detected more pro-Ukraine bots is true, there's also this conclusion which would make a much better headline:
> Pro-Russian non-bot accounts are most influential overall
The Ukrainian bot networks are much bigger but also less effective than the targeted Russian troll farms when it comes to the invasion of Ukraine, at least the way this was measured.
I don't see how they've compensated for the difference in popularity of Twitter in Ukraine and Russia. Russia has its own social media full of its own propaganda, there aren't that many Russians on Twitter in comparison. Furthermore, Russians speaking out against the government can face some serious consequences so the pro-Ukrainian Russians probably know better than to voice their opinion with a hashtag, shifting the balance further.
I also question the reliability of this machine learning network; this data doesn't come from a human verified data set, just from a machine learning algorithm trained on other bot data. I'm sure the network can detect bots that behave like the bots flagged in the original data set it was based on (though I hope that didn't include the faulty French data set) but that doesn't prevent one from setting up a new bot network that this system doesn't catch.
Finally, regarding biases of the linked website, India is generally more in favour of Russia in general. I'm not sure why India chooses to support Russia in such a black-and-white situation, especially given their own history of being invaded by greedy expansionist powers that claim to "help" the local population, but I suppose the dislike for the west is stronger than the compassion for the Ukrainians.