Predicting the past


Digital humanities have a long tradition of using advanced computational techniques and machine learning to aid humanistic enquiry. In this paper, we concentrate on a specific subfield of machine learning called predictive analytics and its use in digital humanities. Predictive analytics has evolved from descriptive analytics, which creates summaries of data, while predictive analytics predicts relationships within the data that also help to explain new data. Predictive analytics uses machine learning techniques but also traditional statistical methods. It uses properties (or features) of the data to predict another target feature in the data. Machine learning is used by predictive analytics to establish the rules that given a certain combination of features make the target more or less likely. Predictive analytics can thus be considered to be a technique to machine-read data. The paper discusses the background of predictive analytics, its use for predicting the past and finally presents a case study in predicting past gender relations in a historical dataset. Predicting the past is introduced as a method to explore relationships in past data.

Digital Humanities Quarterly