A schematic diagram of generating Q-Felt Map. By continuously monitoring the online search queries that contain the keyword “Earthquake” after an earthquake occurred, we can roughly obtain the geographical distribution of earthquake reports from users who felt the ground shaking and wanted to seek exact information online. To reduce the noisy information contained in original queries, and retain effective earthquake reports for drawing the Q-Felt Map, we first design a machine learning method to screen the original queries with weighting strategy. Then, we use the Principle Component Analysis (PCA) algorithm to estimate the direction of the semi-major axis of isoseismal lines, which are finally segmented by a density-based clustering algorithm. Indeed, the Q-Felt Map could be constantly revised with the update of new submitted queries.