The concept of sustainability takes into account a problem of urbanisation which causes a phenomenon of urban sprawl. In order to fight it, the strategy of urban densification is introduced and building upward extensions is one of the methods. However, investigating the feasibility of placing an extension on top requires a significant amount of time and effort from structural engineers. Therefore, this research aims to explore computational tools that would simplify the initial exploration process and make the option of extension more appealing. Theoretical background part of the thesis analyses conditions and limitations for building reuse, upward extension and use of timber. The main impediments that could prevent realisation of such project are found. The practical part of the research aims to explore parametric modelling and machine learning (ML) in order to produce a proof of concept ML tool which predicts how much an existing building can be extended based on its parameters. In order to have a sufficient amount of data, parametric modelling is used. For the first step, information is collected about seven realised upward extension projects in the Netherlands. This data defines the object for the parametric model which is built using Rhino plug-in Grasshopper. One part of the model analyses the existing structures with the goal to establish structural reserve while another part examines extensions. Six parameters defining existing buildings are chosen for iteration, by changing these parameters, a database of 8 652 fictional projects is generated. For the ML, the data is analysed and used for supervised learning where regression task is performed. Multiple linear regression (MLR) and nonlinear regression (NLR) algorithms are applied to make predictions which are validated using 5-fold cross validation (CV) and evaluated by calculating errors. Finally, the research presents a possibility and an example of using ML for extension exploration. The use of ML model during the initial stage of upward extension projects seems very appealing due to the quick feedback with limited information provided. A useful database of realised projects with a selection of structural parameters together with the incorporation of non-structural ones and a customised ML model should provide accurate predictions. However, the current problem of lack of data about realised extension projects needs to be solved.