Vulnerabilities in modern cloud analytics have become more pronounced with the rapid adoption of cloud-based analytics in high impact and sensitive domains like healthcare, finance and IoT; the demand for privacy preserving means that scale to these large-scale problems has intensified. However, existing traditional encryption schemes do not allow processing calculation on encrypted data, enabling the use of Homomorphic Encryption (HE) in analytics workflow. In this paper we propose Hybrid Homomorphic Encryption with Differential Privacy Calibration (HHE-DPC), a novel hybrid approach, that fuses the computational flexibility of leveled HE with the statistical guarantees of differential privacy. The proposed framework provides end-to-end confidentiality with respect to data processing over cloud systems and keeps analytical utility. The effectiveness of HHE DPC is validated with existing privacy preserving methods in terms of accuracy, communication overhead, computational cost, throughput, and convergence. Experimental results show that HHE-DPC exhibits better performance: the accuracy is improved by up to 5.4% and computation overhead is reduced up to 30% compared with competition schemes, including state of the art PPFL-LQDP.