To address the lack of data classification functionality in current homomorphic encryption schemes, this paper proposes an improved RNS-CKKS hierarchical homomorphic encryption scheme. The proposed scheme modifies the modulus generation conditions of the original RNS-CKKS scheme by expanding the selection range of individual prime numbers and introducing computational-level constraints during decryption to restrict the product range of consecutive small primes in the modulus chain. The modified RNS-CKKS scheme significantly increases the number of modulus levels available and ensures accurate scaling factors at specific computational levels. The scheme employs a residue number system (RNS) to enhance computational efficiency and security, ensuring that data can only be encrypted to higher computational levels when a more complete modulus chain is known. The hierarchical RNS-CKKS scheme is implemented based on the Lattigo homomorphic encryption library. The experimental results demonstrate that the proposed scheme maintains stable scaling factors at specific computational levels. Additionally, the accuracy of the hierarchical scheme was confirmed through the experiments.