人格反映了一个人特殊的行为模式，具有较强的稳定性，它影响着人们的行为决策，对其进行预测能够评估人们未来的表现，对于人才选拔等应用场景具有重要意义。近年来，使用脑电信号预测大五人格成为研究热点。然而当前大多数方法都是采用传统的机器学习算法（如支持向量机等），且每次只能预测人格的一个方面，还未出现大五人格整合预测方法，在实际应用中需要多次建模运算才能进行较全面的评估，消耗大量时间和算力，因此对大五人格的整合预测更符合现实应用下人们的需求。

本文采用深度学习算法，构建了一种基于情绪脑电的3D卷积神经网络模型，整合了五个人格预测模型的功能，能够同时对五个人格特质进行预测。实验采集了受试者观看情绪图片时的脑电数据，从中构造出时间、空间、频段三个维度的特征矩阵作为本文所构建模型的输入。该模型通过执行3D卷积操作从三个维度中提取到有关人格信息的更深层特征，最后同时输出五个人格特质的预测结果。为了提高模型的泛化能力，采用留一交叉验证法对模型进行训练和评估。最终，本文建立的模型的分类准确率达到了65.52%。实验结果表明，本文提出的方法作为将深度学习算法应用于大五人格整合预测研究的一个尝试，具有一定的参考价值。

Personality reflects a person's special behavior pattern and has a strong stability. It affects people's behavior decisions. Predicting it can evaluate people's future performance, which is of great significance for application scenarios such as talent selection. Recently, using EEG signals to predict the Big Five personality has become a research hotspot. However, most of the current methods adopt traditional machine learning algorithms (such as Support Vector Machine, etc.), and can only predict one aspect of personality at a time. There is no Big Five personality integration prediction method, which requires multiple modeling operations to carry out comprehensive evaluation in practical application, which consumes a lot of time and computing power.Therefore, the integrated prediction of the Big Five personalities is more in line with the needs of people in practical applications.

In this paper, deep learning algorithm is adopted to construct a 3D convolutional neural network model based on emotional EEG, which integrates the functions of five personality prediction models and can simultaneously predict five personality traits. The EEG data of subjects when viewing emotional images were collected in the experiment, from which three dimensional eigenmatrices of time, space and frequency were constructed as the input of the model constructed in this paper. By performing 3D convolution operation, the model extracts the deeper characteristics of personality information from the three dimensions, and finally outputs the predicted results of five personality traits simultaneously. In order to improve the generalization ability of the model, the leave-one-out cross validation method was used to train and evaluate the model. Finally, the classification accuracy of the model established in this study reached 65.52%. The experimental results show that the method proposed in this paper has certain reference value as an attempt to apply deep learning algorithm to the integration prediction of the Big Five personality.