Event sequence analysis is a vital area of research in data mining and machine learning, with applications in social networks, financial transactions, and medical monitoring. Understanding causal relationships between events helps predict user behavior and market trends, providing interpretable decision support. However, existing methods mainly focus on temporal-domain features and struggle to capture frequency-domain ones. Additionally, they lack robustness to noise, limiting their ability to analyze complex event sequences. To address these challenges, we propose a novel model, the Hybrid Mamba Hawkes Process, to improve causal discovery in event sequences. First, we design a Frequency Mamba (FMamba) Block by introducing the Fourier and Wavelet Transforms into the Mamba model. The Fourier Transform extracts global frequency features and reduces noise, while the Wavelet Transform captures local frequency characteristics, enhancing the model’s ability to track event dynamics. Second, we design a multi-domain feature residual block that integrates temporal-domain, frequency-domain, and raw data features. By leveraging attention mechanisms and multi-layer perceptrons (MLPs), the model effectively captures complex causal relationships. Experimental results show that our model outperforms existing methods on three public datasets, which demonstrates its capacity on capturing multi-scale periodic patterns and the dynamics of events.