Machine Perception of Political Manifestos in Predicting Performance of Public Office Holders

Abstract

Electorates expect politicians seeking public office to make known in advance, their intended programs in form of proposal. This is usually presented in speech in form of manifesto. Times within number, manifestos have always precede voting proper whereby the electorates evaluate politicians based on their manifestos. While intention is socially difficult to measure, this study adopts artificial neural network machine learning approach to map-measure the manifestos of politicians and their eventual performance in office. Due to changes in political names and structure, the study could only utilized the manifesto data of the two most popular political parties in Nigeria, from 2007 to 2019. The result of the empirical analysis shows that the model evaluation accuracy stood at 67%. With more adequate data, this result can be improved upon by subsequent research work.