Download

Download Full Text (1.4 MB)

Abstract

Despite the popularity, we noticed that it is rather hard to verify a NLP/text-mining like stock prediction model's performance due to the amount of "groundwork" needed. It is very typical a researcher will have to gather the plain text data, the company info, the stock market data, and categorize them in a way that is communicable with each other and the model; then the researcher will need to build a virtual trading platform that keeps track of all the trading signals generated by the model, log the activities in a certain way, then do some kinds of visualization for evaluations. To implement all these steps from ground up, it is required for a researcher to have certain level of proficiency on skills which are, from a research stand-point, fairly deviated from the nature of the NLP/text-mining model itself (like scraping a website and understanding the fundamental mechanism of trading in stock market). Thus, we like to build a set of lightweight tools that may automate such process to a certain degree.

Symposium Date

Fall 12-1-2012

Keywords

backtest, stock market, NLP, model utility, AI, finance--data processing

Creative Commons License

Creative Commons Attribution-NonCommercial 4.0 International License
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License

An All-In-One NLP Stock Market Backtester

Share

COinS