The accurate determination of thermal conductivity of concrete is very important for the temperature crack control of concrete structures in cold region. Based on the orthogonal experimental method, the variation of thermal conductivity of concrete under the conditions of water-cement ratio, temperature, saturation and volume fraction was analyzed. In order to fully excavate the intrinsic structural characteristics of high dimensional data of thermal conductivity, an exploratory data analysis (EDA) method, PPR non-hypothetical modeling technology, was proposed to analyze and model the experimental data. The stability of the model was evaluated by the method of “ precision consistent test ” which was consistent with the data “ reduction fitting ” and “ reserved test ”, The selection criteria of PPR model samples were put forward through analysis. Finally, based on the established PPR model, the thermal conductivity of concrete was simulated and predicted, so that the thermal conductivity of concrete can be studied more comprehensively. The results show that this modeling method avoids the problems of subjective assumption and uncertainty of solution results, which exist in the confirmatory data analysis (CDA) method, and has good discipline versatility. The research results not only provide a comprehensive theoretical basis for the accurate calculation of the internal temperature field of concrete structures, but also provide effective solutions for the optimization design analysis of various disciplines.