Gaming sentiment analysis

I analyzed about 61,000 tweets about CS:GO and other games to compare player sentiment and find recurring complaints, framed as a hypothetical marketing brief for Valve. Cheating, streaming, and cosmetic items stood out as topics worth investigating.

A multilayer perceptron classified sentiment with 94.7% accuracy in the original analysis. That score describes the classifier; whether a campaign based on it would work would need its own test.