This study aims to identify black hole merge signals by analyzing gravitational wave data with machine learning algorithms. In this research we use two kinds of data, observational data and simulated data I generated. The observational data, including 64 widely recognized samples, is from the Laser Interferometer Gravitational-Wave Observatory (LIGO). The simulated signals are generated base on various black hole masses and signal-to-noise ratios. The simulated signals were mixed with different types of noise and then processed through intensity adjustments, Q transform, and data whitening. Statistical features were extracted from the processed data and used to train eight different machine learning algorithms. Results indicate that the decision tree algorithms outperformed the other algorithms in predictive accuracy. In addition, the standard deviation of intensity difference is identified as the most crucial feature, showing an importance score of 27%. Our model demonstrated high efficiency in gravitational wave signal identification and successfully reduced model complexity, making it more suitable for practical applications.