Abstract: Total organic carbon (TOC) mass fraction is an important index of source rocks evaluation. In order to evaluate the organic carbon of source rocks in Chang 9 Member of Yanchang Formation in Ansai area, southeast Ordos basin, firstly，this article  establishes w(TOC) models for quantitative prediction of well logging by applying the multiple regression model, the traditional Δlog R model, the improved Δlog R model and the generalized Δlog R model, based on core analysis of measured w(TOC) data and  the response characteristics of source rocks to different logging curves.  Secondly, by   analyzing and combining  these models,  the fitting superposition coefficient  extracted from the improved Δlog R model is applied  to the calculation of two generalized Δlog R models, and the application effect is  good. Finally, the four models are compared and optimized, and the most suitable quantitative prediction model for source rocks in the study area is proposed. The results show that the generalized Δlog R model considering the density factor has the highest accuracy, with an average relative error of 7.78%； The multiple regression  model   has the second highest accuracy, with an average relative error of 9.65%. Both of them can meet the accuracy requirements of quantitative prediction of w(TOC).

Key words: logging prediction method, Δlog R, organic carbon, source rocks, multiple regression, Ordos basin, Yanchang Formation