Evaluation of heavy metal pollution in soil mass relies mainly on index methods. Choosing suitable methods to process indexes is key to obtaining accurate and reasonable evaluation results. Connection cloud models have been commonly used to evaluate pollution indexes. However, generating a cloud at each grade for each index is a tedious calculation. To address the randomness and fuzziness of soil heavy metal pollution evaluation indicators, we propose a comprehensive connection cloud model that couples the connection clouds of all indicators at each grade as sub-clouds. We establish the model by combining the weights of each indicator to determine the soil heavy metal pollution grade. The proposed model is effective and feasible for the evaluation of soil heavy metal pollution, as demonstrated through examples and comparisons with other methods. The comprehensive connection cloud model overcomes the traditional cloud models’ defect of requiring normal distribution for evaluated indicators and considers the interaction between indicators. Thus, it provides a novel route for uncertainty analysis, considering the fuzzy randomness between multiple interrelated indicators.