Harnessing Machine Learning to Decode YouTube Subscriber Dynamics: Regression Predictive Models and Correlations

Abstrak

YouTube has grown and become a digital media giant. Content creators continue to struggle with predicting subscriber growth. Due to viewers' changing interests and the vast amount of information, it is challenging to determine which factors most influence subscription behavior. Optimizing content strategy and ensuring channel growth need an understanding of these traits. This study uses linear regression models (LR), neural networks (NN), and Gaussian processes (GP) to predict YouTube subscribers and examine category correlations using video data from various topics. The study of correlation matrix analysis was performed with an absolute root mean square error (RMSE) of 26256351, and the NN prediction model outperformed the LR and GP models. The correlation matrix indicates a slight positive correlation of 0.067 among the YouTube categories. Specifically, the correlation coefficients for population, unemployment rate, and urban population are 0.080, -0.012, and 0.082, respectively. These findings suggest future research to create more intentional content and search for significant factors that increase viewership and marketing audience growth.