Cloud computing is an information technology model that provides computing and storage resources as a service. Data storage security remains the main challenge in adapting this new model. The common solution to secure data in the cloud is data encryption. However, handling all the data with the same security policy does not appear to be good practice, because they do not have the same sensibility for the data owner. The present research proposes a new method to improve the security of data in cloud storage. It combines two domains represented by machine learning and multi criteria decision making, in order to provide a new classification method, that classifies data before being introduced into a suitable encryption system according to their category. A Cloudsim simulation has been used to demonstrate the effectiveness of the proposed method. The results of the simulation exhibit that our method is more efficient and accurate and takes less processing time, while ensuring data confidentiality and integrity.