search

philipperemy / financial-news-dataset

Reuters and Bloomberg
222 stars 95 forks source link

Various Questions #1

Closed xiao666 closed 6 years ago

philipperemy commented 6 years ago

Hey! Where do you see the pickle files?

xiao666 commented 6 years ago

im doing my thesis on NLP application on stock prediction, just searched related dataset, and find here : )

philipperemy commented 6 years ago

Oh yeah you're referring to the Reuters data set?

github.com/philipperemy/Reuters-full-data-set

xiao666 commented 6 years ago

yes, the problem was the python version. I can load the headlines now. just a bit more questions about the details of news dataset,

  1. each day may consists of different number of news right?
  2. and if I want to use these news to make stock prediction, how to align the news with market performance together? depend on the same date?
philipperemy commented 6 years ago

Each news has a precise datetime.

For example this one:

ts = 20070228 11:46 AM EST, t = European stocks hit 7-week low amid new sell-off, h= http://www.reuters.com/article/companyNewsAndPR/idUSWEB277620070228

was published on February 28th 2007 at 11:46 AM (Eastern Time Zone).

You can align the news with the market performance by:

Or if you have 1 min price data, you don't need to aggregate.