This study utilizes a brain—computer interface (BCI)—based deep neural network (DNN) and genetic algorithm (GA) method. This research explores the interaction design of the main control human-machine interaction interfaces (HMIs) for intelligent electric vehicles (EVs) by integrating neural network predictions with genetic algorithm optimizations. Augmented reality (AR) was incorporated into the experimental setup to simulate real driving conditions, providing participants with an immersive and realistic experience. A comparative analysis of several models including the support vector machines-genetic algorithm (SVMs-GA), decision trees-genetic algorithm (DT-GA), particle swarm optimization-genetic algorithm (PSO-GA), and deep neural network-genetic algorithm (DNN-GA) was conducted. The results indicate that the DNN-GA model exhibited superior prediction accuracy with the lowest mean squared error (MSE) of 0.22 and mean absolute error (MAE) of 0.31. Additionally, the DNN-GA model demonstrated the shortest training time of 69.93 s, making it 4.5% more efficient than the PSO-GA model and 51.8% more efficient compared to the SVMs-GA model. This research focuses on promoting an innovative and efficient machine learning hybrid model with the goal of improving the efficiency of the human-machine interaction interfaces (HMIs) interface of intelligent electric vehicles. By optimizing the accuracy and response speed, the aim is to enhance the control interface and significantly improve user experience and usability.