<![CDATA[Artificial intelligence is a rapidly growing technology that has been implemented in various sectors such as finance, education, the military, healthcare, and business. One form of artificial intelligence is deep learning, which has the capability to perform complex pattern recognition and decision-making using very large datasets. This capability makes deep learning one of the key technologies adopted by companies and organizations to improve their policies and enhance operational stability. However, several studies have shown that deep learning still faces significant challenges, particularly security risks that may expose sensitive business or organizational information. Therefore, in this work, we aim to address this problem and enhance the privacy protection of deep learning by incorporating differential privacy. Differential privacy is a technique that protects sensitive information by adding controlled noise to the data or model outputs, thereby reducing the risk of information leakage. We evaluate our proposed method using marketing data and implement it to 5 different models, and based on the experimental results, Our proposed method achieves its best performance using the linear regression model, yielding an RMSE of approximately 1821.30, an MAE of 1406.45, and an R² of 0.007717 for higher privacy budget. Under a lower privacy budget, the performance of the linear regression model decreases to an RMSE of 3634.24, an MAE of 2799.05, and an R² of –0.008482, yet it still outperforms the other four model approaches.]]>
Copyrights © 2025
