Online Portfolio Selection Based on Investor Sentiment in Stock Microblogs
児矢野晋太 (1651049)
Stock microblogs are social media where various opinions on stocks are talked about and users can reveal their sentiment for stock price movement.
In this work, we used the sentiment to address online portfolio selection problem, where the objective is to maximize cumulative return by sequentially updating a portfolio.
For that purpose, we took two approaches in a stepwise manner.
The first one was to estimate missing sentiments from the user's post in order to increase available ones since few users reveal them.
The second step was to construct a portfolio by following reliable user's sentiment and opposing unreliable user's one, where the reliability was calculated based on the past sentiment and the stock price movement.
As the result, the proposed method estimated sentiments with high accuracy and got the largest cumulative return among existing methods.
It was shown that the proposed method has the competitive performance, but more effective ways to exploit unreliable users need to be explored.
Your_MS_Thesis_Or_Report_No.(例：NAIST-IS-MT*******): Your Name