Abstract: Under the background of big data, the continuous growth of geological data poses challenges to traditional discrimination diagrams represented by TAS: On one hand, the excessive data points within limited diagram space reduce readability and hinder effective visualization; On the other hand, the input of new data into traditional diagrams with outdated original data may lead to perturbation in classification boundaries, compromising the stability of discrimination results and compatibility with existing literature plots. To solve the above problems, this study first extends the previous research on TAS diagrams by constructing category partitions based on spatial positions for various lithology labels in classic diagrams. The discrimination is made then on the basis of spatial relationship between location of the data to be classified and category partitions, with results presented in data table to mitigate readability degradation caused by the increase of data volume. Besides, over 240 000 entries of major element data of igneous rocks are extracted from GEOROC database for TAS visualization and for kernel density analysis in terms of lithologic classification. The corresponding category probability field is constructed across the plotting coordinates based on the analysis results. The probability is calculated by the position of the data to be classified in each probability field, and the probability results of different lithologic labels are compared. Based on probability field, the known lithology label data are used to distinguish the data to be classified, supplement the traditional classification boundary model and form more quantitative discrimination results.