Abstract: In order to study the transparency and similarity requirements of transparent soft soil simulants, transparent cemented soil was synthesized using fused quartz sand, nano-silica, n-dodecane, and 15# white oil as base materials, guided by an orthogonal experimental test. Key physical and mechanical parameters (unit weight γ, internal friction angle φ and cohesion c) were quantified through ring knife tests and consolidated fast shear test. Combined with multiple linear regression analysis and support vector machine evaluation model, the optimal mix ratio integrated design method of transparent similar materials was developed. The results show that the synthesized transparent cemented soil replicates the physical-mechanical properties and compression behavior of natural soft soil. Based on the multivariate linear regression equations and support vector regression model obtained by orthogonal test data fitting and machine learning, an integrated design method of optimal mixture ratio and geometric similarity constant is established.