In the context of big data, simulation models often involve multiple factors. The purpose of simulation screening experiments is to identify the small subset of factors(also referred to as simulation inputs or variables) that have the most significant impact on the response(simulation output or system performance). Currently, two commonly used screening methods are sequential bifurcation(SB) and elementary effects(EE). In recent years, the EE has gained traction in various fields due to its advantage of not assuming specific mathematical relationships between simulation inputs and outputs(i.e., it is model-free). However, it suffers from computational inefficiency. To address this issue and improve the computational efficiency of the EE, this paper proposes an enhanced and more versatile method called enhanced cluster sampling(ECS). Unlike existing cluster sampling methods, the ECS automatically constructs a sampling matrix by decomposing the matrix, enabling the generation of an equal number of target elementary effects for each factor. This approach significantly saves simulation budget. Monte Carlo simulation experiments demonstrate that the ECS greatly enhances computational efficiency without compromising statistical effectiveness, providing compelling evidence for the effectiveness of the proposed method.