Problem Statement
In the field of portfolio and investment theory, there is currently a debate on whether or not technical analysis, and momentum trading in particular, are rendered obsolete by perfect market efficiency. This project applies machine learning techniques to attempt to contribute to this debate.
The implications for this project are important not only because of its application to investing strategies, but also because of its contribution to the debate on the efficient market. Our eventual conclusion takes a side in this debate and supports it with data obtained through this project.
Why Machine Learning?
Price and volume information vary greatly among different stocks. The goal of this project is to see if momentum trading can be generalized and hold true across the entire market. In order to make an accurate prediction of future trends, price and volume data for each stock must be either reduced to a standard scale (which is very hard to do), or a function can be fit to all of the data points such that a penny stock can be inputted into the same formula as BRK.A.
In order to make this function as accurate as possible, the importance of each attribute must be determined, as some pieces of information about the stock are more valuable than others -- it is unlikely that all input information is equally important to the future trend of a stock (the returns between two weeks a year ago is likely less important than the return between two weeks one month ago). In order to compute these weights, a machine learning algorithm that is able to learn which attributes contribute the most is necessary.More details about this model can be found under the Model Creation tab.
In the field of portfolio and investment theory, there is currently a debate on whether or not technical analysis, and momentum trading in particular, are rendered obsolete by perfect market efficiency. This project applies machine learning techniques to attempt to contribute to this debate.
The implications for this project are important not only because of its application to investing strategies, but also because of its contribution to the debate on the efficient market. Our eventual conclusion takes a side in this debate and supports it with data obtained through this project.
Why Machine Learning?
Price and volume information vary greatly among different stocks. The goal of this project is to see if momentum trading can be generalized and hold true across the entire market. In order to make an accurate prediction of future trends, price and volume data for each stock must be either reduced to a standard scale (which is very hard to do), or a function can be fit to all of the data points such that a penny stock can be inputted into the same formula as BRK.A.
In order to make this function as accurate as possible, the importance of each attribute must be determined, as some pieces of information about the stock are more valuable than others -- it is unlikely that all input information is equally important to the future trend of a stock (the returns between two weeks a year ago is likely less important than the return between two weeks one month ago). In order to compute these weights, a machine learning algorithm that is able to learn which attributes contribute the most is necessary.More details about this model can be found under the Model Creation tab.