Local Differential Privacy (LDP) has advantages of relying on no trusted third parties. However, if the sensitivity of data is not considered, which will lead to imbalanced privacy protection, lower data accuracy, and lower utility. For privacy sensitivity differences among the data, we propose an optimized algorithm called Sensitivity Dynamic Difference Report Mechanism (SDDRM). In SDDRM algorithm, the dynamic difference report mechanism is adopted to allocate different privacy budgets for sensitive and non-sensitive data, which can provide stronger privacy protection for sensitive data and effectively reduce noise interference for non-sensitive data. In other words, the accuracy of data is improved while the data privacy is satisfied. Finally, experiments are conducted on two real datasets, Stocks and SynB, and an optimal selection method is given through different Settings of privacy budget allocation, which can achieve the balance between data privacy protection and data accuracy. Comparative experiments show that the loss indicators are significantly reduced by the optimized SDDRM algorithm (compared with the DDRM algorithm, the percentage of reduction is between 5% and 45%), which confirm that the optimized SDDRM algorithm can improve the data accuracy while protecting the data privacy.