This repository was created, to host my Bachelor's Degree final project. I will mainly focus on the creation of deep learning architectures, metastrategy composed of several deep learning models operating through a voting procedure, to efficiently predict financial products prices movements. I have collected closing, low and high prices from 2000 to 2021 for each financial product that interested my research. The daily observation is labeled as buy (1) or sell (0) given the price of the following day; if the closing price of the t observation is lower than the closing price of the t+1 observation, we will label the t observation as buy (1) and viceversa. Technical indicators, correlated financial products, and a modeled time window are adoperated to create a 3D tensor which corresponds to the "image" associated to each observation. Subsequently, the deep learning architecture is deployed to predict the label of future observations.
matteo-ticli/stock_prices_classification
Bachelor's final project. Financial Indexes movement classification via Deep Learning.
Jupyter NotebookMIT