Using News to Predict Stock Movement (Student Presentation, Group 12)

This is a STAT 451 class project presentation

by Chuyang Chen, Chutong Jiang, and Xinjie Lu

This presentation is shared with the students' permission.

Abstract:

We use machine learning algorithms to predict U.S stock movement based on the number of financial news. The textual data nowadays plays a very significant role in making predictions for the stock market. However, due to the complexity of the stock mar-ket, it’s hard to make precise and accurate predictions based on the context of the news. Also, we found that there might be a strong correlation between the number of news and the stock movement: usually when a big event happens, the related news would be published and the stock may have dramatic changes. Hence, we choose 37 tech firms with market values from 30 to100 billion dollar as our stock data and corresponding100 influential and financial news as our news data from Finviz’s website. We divide our dataset into a70% training set and 30% testing set. The models we use include K-Nearest Neighbors algorithm (kNN) and Random Forest. By selecting the optimal hyper-parameter for the model, the accuracy of them is in-creased. Both of them perform quite well so this approach is likely to be used in real world applications to give some insight and direction for those who intend to invest in the stock market.