Increase in utilization of multimedia enhanced the risk of authorized and secured data, because data is growing exponentially and possess complex operations. Currently, digital image watermarking is the technique to make digital data more authentic and protected. The main objective of using image watermarking is to stabilize the factors like capacity, security, robustness and authentication of watermark against certain attacks. In this model, a machine learning approach for optimizing the major coefficients and factors during image watermarking is applied through the concept of ensemble learning. Firstly, original image is transformed according to DWT by decomposing the original image so as to embed the watermark and then SVD is applied to embed the watermark into their equivalent singular values. The purpose of optimization is to reduce the noise content and increasing the strength of several image watermarking factors by using the ensemble classifier. This method provides better results with respect to MSE, PSNR and NC in comparison with existing system.