US-Sector-Forecasting-with-Machine-Learning

We designed an algorithm for predicting the behavior of US Sector ETF prices, as a proxy for the sectors themselves, to function as an asset allocation framework. Using API calls from existing frameworks in R, we mined monthly price data for a myriad of assets that best represented broad market information to use as features in modelling the sector ETFs. We tested mutliple machine learning techniques to cluster like asset classes and select only the most significantly correlated from each group as components for each model. We then used both linear and decision tree models to predict future prices. All combinations of filters and models were evaluated against one another for accuracy and error rates.