Classificare fondi d'investimento con metodi di apprendimento automatico

In the world of financial instruments, funds are widely used tools that possess particular characteristics that make them uniquely interesting for both saving and investment purposes. One of the main drawbacks is the fact that, due to these very characteristics, funds are extremely diversified, thus for investors finding the most suitable opportunity is a difficult and lengthy process. Firms offer a possible solution to this problem through fund classification, a task that sees financial analysts individually assigning funds to classes, collections sharing certain characteristics, making the world of funds easier to navigate for the investors. Since this process is expensive and time-consuming, help might come from machine learning, the field of computer science that employs algorithm and model to solve tasks without needing specific instructions, but relying instead on pattern recognition and inference. For this purpose, the goal of this thesis is to search for a machine learning pipeline that might help with the classification of investment funds, in the specific setting of the fund database provided by the financial data analysis firm FIDA.\\ To this aim, we'll conduct a project that stems from a theoretical overview of funds and the classification problem, to the exploration of our data and data management, ending with the application of a series of machine learning methods and with the presentation of techniques for optimizing the performance. Developed with support from FIDA, this thesis aims at being the stepping stone for a project that will lead to the employment of machine learning-based methods as a key component for the business of automated financial counseling.