Machine “ Gaydar ” : Using Facebook Profiles to Predict Sexual Orientation

The rise of social media and the large, rich datasets they make accessible have allowed us to learn about people not only through information shared explicitly, but also through information shared implicitly in the form of trends and patterns. Here, we emphasize the value of implicit data by creating a machinelearning algorithm that uses basic information, photos, and published text on Facebook profiles to predict sexual orientation in males. We constructed a model with Naïve Bayes classifiers and a Support Vector Machine, performing on different types of data. We used 10-fold stratified cross-validation on our dataset as a measure of generalization error. Currently, we have created a model with an accuracy of 91.02%.