Application of Deep Learning for Level of Engagement in Civic Activities Prediction in Emerging Adulthood for Smart City Development
We applied the deep learning method, which has been developed in the fields of computer and data science for accurate prediction, to predict political purpose development during emerging adulthood. We tested whether deep learning more accurately predicted Wave 2 political purpose with Wave 1 predictors compared with traditional regression. A convolutional neural network consisting of two dense and dropout layers was trained to predict the outcome variable. For comparison, we also estimated a multinomial logistic regression model. The result demonstrated that deep learning outperformed traditional regression in general while effectively minimizing overfitting. Moreover, from exploratory analysis, we found that deep learning might be able to model the non-linear relationship between the predictors and outcome variable. Based on the findings, we discussed the implications of the present study within the context of improving citizens’ lives in smart cities.