The invention discloses an intracardiac abnormal activation point positioning model construction method based on CNN and LSTM, and the model can achieve the good positioning of the specific position of a VT abnormal activation point under the condition that the 12-lead body surface potential data of a patient is obtained, and obtains the three-dimensional coordinates of the position. According tothe invention, the idea of deep learning is introduced into ventricular tachycardia abnormal excitation point positioning; in the training stage, the collected QRS data is used as input; the three-dimensional coordinates of the QRS data corresponding to the mapping points are taken as labels to train a CNN-LSTM network, feature extraction is performed on the input data by using Conv1D, feature fusion is performed on a time domain by using LSTM, regression prediction is performed on the three-dimensional coordinates by using a full connection layer, and finally the CNN-LSTM network is constructed. According to the network model, the position prediction of the VT abnormal activation point is realized from the perspective of data driving, and the time-consuming and labor-consuming problems ofcatheter ablation in clinic are effectively solved.