The Cell-Free Massive Multiple-Input Multiple-Output (CF-MIMO) system is a promising technology for beyond-fifth-generation (B-5G) networks. It involves deploying multiple access points (APs) with multiple antennas to serve groups of users (UEs) cooperatively. In this paper, we introduce two algorithms that utilize big data technology for interference identification and signal-to-interference-plus-noise ratio (SINR) prediction. These algorithms effectively identify user-level interference information and provide support for resource allocation. They outperform traditional machine learning methods in terms of accuracy, time efficiency, and computation complexity by a significant margin. To establish the theoretical foundation, we derive the closed form of the average SINR based on large-scale fading coefficients (LSFCs). Our results demonstrate that our algorithms significantly enhance prediction accuracy by 50%-75% and achieve an impressive 8-to 80-fold improvement in training efficiency compared to the benchmark scheme.