Predicting game ownership dynamics: a novel POAFD-trend analysis approach

This article introduces a pioneering method for predicting video game ownership utilizing pre-orthogonal adaptive Fourier decomposition (POAFD)-trend analysis. Predictive modeling in well-established domains relies heavily on data quality and historical trends, whereas game sales prediction faces unique challenges due to limited and uncertain data. To address these issues, we develop the POAFD-trend theory, which decomposes data signals into a trend component and its complementary parts. By combining statistical analysis with the stochastic POAFD algorithm, we achieve highly accurate forecasts for similar game categories. Extensive experiments on real-world data demonstrate that our approach outperforms state-of-the-art methods, achieving an \(\textrm{R}^2\) ${\text{R}}^2$ score of 0.57 and low error rates (MAPE: 0.33, MAE: 7148) on Detroit Become Human sales predictions. Our method effectively mitigates challenges posed by limited data, significant randomness, and high uncertainty, thereby advancing the field of game ownership forecasting. Our project website is at https://github.com/CXLao/POAFD-trend.