Epilepsy patients typically rely on resection surgery of the epileptogenic zone (EZ) to control seizure recurrence, and precise EZ localization is crucial for surgical success. Previous localization algorithms primarily relied on temporal features of single-channel electroencephalogram (EEG) signals, overlooking interactions between brain regions and resulting in insufficient utilization of spatial information. Based on stereotactic EEG (SEEG) data of 23 epilepsy patients, this study applies a one-dimensional convolutional and long short-term memory hybrid model (Conv1D-LSTM) to fuse single-channel and multi-channel features for EZ localization. The method first extracts time-domain, frequency-domain, and nonlinear features from single-channel SEEG signals. Then, it constructs brain functional networks during epileptic seizures based on phase locking value (PLV), transfer entropy (TE), and generalized partial directed coherence (GPDC), respectively. Graph theory features are extracted from each network, and statistical analysis is conducted to determine whether significant differences exist between epileptogenic and non-epileptogenic channels across the aforementioned features. Subsequently, these features are applied to the Conv1D-LSTM to integrate spatiotemporal information for EZ localization. Experimental results demonstrate that compared to using either single-channel or multi-channel features alone, the Conv1D-LSTM based on the fusion of both features achieves a localization accuracy of 93.31% on the SEEG dataset. Ablation experiments show that the graph-theoretic features of brain networks play a more significant role in the localization of EZ than single-channel features, and the dual-attention mechanism significantly improves the localization performance. Comparative experiments show that Conv1D-LSTM outperforms mainstream models such as support vector machine, convolutional neural network, temporal convolutional network, and bidirectional LSTM. This study provides a novel auxiliary tool for the precise EZ localization in clinics.