Mobile devices are gaining impact in last few years due to their versatility of platform to run various types of applications. Over the years, the storage of mobile devices is increased but still it cannot fulfill all users requirements. Most of the times, users faced limited memory issues on mobile devices, therefore to free up space users often delete files abruptly without considering the usage requirements in future. Thus, as a result users are unable to find important files incase of no internet access. In this paper, we proposed a recommendation system to facilitate users to upload less frequent and important files to cloud storage and thus, create space for new data files. Our proposed model is trained on users activities, two different classifiers are implemented and results are compared. The evaluation section shows that using simple classifiers we have achieved an accuracy of 58.97 percent. At the same time it has been observed that due to limited computational resources on mobile devices it is not feasible to actively train the model to suggest appropriate files for offloading. Therefore, to implement more complex algorithms, and to cater limited computation problem, we proposed a architecture to train our models at cloudlets and download on client mobile devices on availability of internet connection.