Welcome back to our blog, today we are going to see how to Model a program that reads news from websites and evaluate their capacity to affect (or represent a popular perspective on) the title/stock/option.
First of all we want let you know that we own the copyright on this posts, therefore this is purely informative to satisfy your curiosity and you can't steal this methods without our license, if you are interested connect with us.
The first things we need are these:
1- A list of sources where to get reliable and early news (and their value of reliability)
2- A list of words positively and negatively related to the Title/option (and their traslation)
3- A library of words linked to a positive or negative meaning and value (and their traslation)
The program, again, is based on simple math and simple coding, the first thing it does is to split an article in Title-SubTitle-Text (or Body), in terms of effectiveness on the Stock/option each of this subsections are weighted differently, in this order Title-Body-SubTitle.
The text the is splitted, by dots, into other pieces called Periods at this point inside every Period the words that belong in the previously declared (in the lists #2 and #3) are then translated into the listed value and placed in the Period sequence, that, by default is an empty string that, by translating the words we need gets filled with values.
Note that the program translates only the words in the lists in order to avoid to translate junky informations and words that are not relevant.
Translating those words let us exploit them only by their meaning, in terms only of their value and not their form, reducing them to their essence.
At this point the sequence of each period, once compressed, in order to make it shorter and ordered and this result, just like for all the other periods will be placed into the Body sequence that works in the same way as before, and again compressed.
At this point as we should have a final value for the text, title and subtitle (which are processed in the same way) are all inserted in a temporary sequence and compress it once again (keeping in mynd the hierarchy and therefore that the title's values counts more than the text's).
The final stage is then to write the final conversion for the article, this is done by writing a sequence that starts with the value of the source (in the #1 list), followed by date and hour of the new and finally the value we obtained before by compressing the overall sequence.
Another feature of this program is that it can also take an average value with news from other sources if they've been realed early enough.
This is it for today,
StatsApplied