Analysis and development of algorithms for cloud and smart city

This thesis is concerned with cloud computing and big data for smart cities. As far as the cloud is concerned, a framework has been developed, which can create patterns relating to the workload of a virtual machine for a certain period of time and for all the resources that are considered useful during the simulation phase. Using these patterns, it was possible to simulate the workload of a datacenter and nd the best allocation of its virtual machines using heuristics to solve the Vector Bin Packing problem. All phases, i.e., the insertion of the datacenter's characteristics, the simulation of the datacenter and the visualization of the results are supported by a webapp. As far as smart cities are concerned, an application has been developed that uses the model and tools made available by the Km4City framework. Based on the application architecture, a \Sii-Mobile Mobile App Develop- ment Kit" has been created, which allows other developers to create their own module to be integrated into the already developed application. A machine learning algorithm has been developed, based on data relating to parking lots with controlled access (large paid parking lots with bar), which carries out daily training of a \Bayesian Regularized Neural Network", which allows to generate predictions (every 15 minutes) of how many free spaces will be available in a specic parking lot, one hour after the prediction has been made. This algorithm has been successfully implemented within the application to make the data related to the predictions generated available to users.