The invention provides a well-seismic data extraction method based on Euclidean distance approximate matching. The method comprises the following steps: carrying out earthquake attribute approximate matching, searching a coordinate point position nearest to a target well, extracting an earthquake attribute value corresponding to the position approximately as an earthquake attribute value corresponding to the position of the target well, and integrating all earthquake attribute values into an attribute set; carrying out lithologic data approximate matching, utilizing time-depth conversion to convert a time range into a depth range corresponding to the target well through the beginning and ending time range of a time window, and therefore, calculating a lithologic proportion in the depth range corresponding to the target well sample; and matching the extracted earthquake attribute set with a lithologic type to obtain an earthquake attribute set with a category tag. By use of the method,the position of the target well is approximate to a nearest grid position in adjacent distance, the well-seismic data extraction of the target well is realized, the numerical category tag and the earthquake attribute value can provide sample point support for the training of a machine learning classification model.