Voting intention prediction based on lifestyle and socio-demographic data: a machine learning application

The field of politics plays a significant role in our daily lives, shaping our actions and decisions. Understanding voting intention has been a crucial aspect of studying electoral behaviour, considering various factors that influence it. Among these factors are socio-demographic data, encompassing personal details of voters, and lifestyle data, reflecting their habits. With the advent of machine learning models and their remarkable accuracy, these data types can now be analyzed within a robust scientific framework. This research compares different machine learning prediction models to identify the most effective approach for managing and analyzing such data. To facilitate this comparison, the Datagotchi application provides a dataset that combines socio-demographic and lifestyle data, enabling the application and evaluation of diverse classification models. A comprehensive data preprocessing stage is employed to ensure efficient model application. Through this methodology, specific models will be emphasized based on their accuracy, handling of variables, and prediction of different political parties within the dataset. The results indicate that ensemble classifiers, significantly Boosting, possess very good predictive abilities, surpassing the performance of individual models. Moreover, DT models demonstrate both good accuracy and user-friendliness, whereas KNN encounter difficulties when dealing with these data types. This research offers valuable perspectives on the advantages and limitations of different classifiers in forecasting voting intentions based on socio-demographic and lifestyle data.