In real criminal cases, the decision outcome is often influenced by many complex factors, such as the importance of initial evidence and the prioritization of evidence. How to model these information in an integrated manner to provide technical tools for case detection so as to find the real suspect is of great importance for social security and stability. To address the above issues, this paper proposes a novel soft likelihood function based on the Decision Making Trial and Evaluation Laboratory (DEMATEL) method. Firstly, the proposed method well preserves the preference of decision-maker (DM) in the soft likelihood function proposed by Yager et al. Secondly, the method takes into account the modeling of associated information. In addition, it also extends the soft likelihood function to reflect the preferences of DMs through the importance of evidence. Finally, based on these designed algorithms, a decision processing model for criminal cases is constructed, which systematically provides a guiding process for case detection. Numerical examples and applications show the practicality as well as effectiveness of the proposed method.