Mind the Gap: Understanding Gender Inequality in Movie Industry using Social Network Analysis and Machine Learning

Gender inequality in the movie industry has increasingly received attention in recent times. This paper aims to understand gender inequality in the movie industry using social network analysis and machine learning by predicting whether a movie would pass the Bechdel Test. In particular, social network data constructed from movie subtitles to indicate the interactions between characters is used, in conjunction with the movie metadata to characterize each movie. Results from social network analysis showed that female actors generally occupy less important social positions and play a less important role in the social networks in almost all genres. This is even prevalent in Romance and Family movies that basic statistics find more female actors, suggesting that social network analysis provides non-obvious values using traditional methods. Moreover, the gradient boosting tree model can achieve 77.4% of accuracy to correctly classified a movie to be gender equitable or not, with the help of the graph features obtained from the movie social networks. This is 16% uplift compared to using only the movie metadata. Overall, social network analysis not only provides new findings of female actor's role in the movie, but also can greatly help machine learning model to achieve much better performance to predict the gender disparity in movies.