Frequent subsequence compression

[Hassanbenlebsir]

Hi,
Brilliant work your article is really compréhensive and you made everything so clear.

i'm working on a subject and i'm coding CPT in python.

subject describtion :
my problem is based on the past for instance this is my past A-->B
I want to know the out come for this past.
so i look at my tree and i search for all sequences with the same past and see the its outcome .
example: A-->B-->C so for this case C is the outcome for this past (that means C+=1) and i do this for the whole tree
so i can suggest the outcome for the past A-->B .

until now i can do this with CPT and i want to improve it and go to CPT+ using FSC.
I followed your steps i calculate the frequency of every subsequence and i fixed a minsub

questions:
My first question:
How to fix this minsub ?
second question FSC:
For this this question i'll give an example first
ex:we have this path A-->B-->C-->D and we know that the B-->C is frequent so i replace B-->C with X : A-->X-->D.
so now when i want to calculate the outcome for my past for example A-->B
the subsequence A-->B it will not be found in this case because B is a part of X so is there away to avoid this?

I'm looking forward to your reply

[philfv]

Hi,

1. For the first question,  the threshold must be set empirically for each dataset.  If you set it too high there will be no frequent sequences.  If you set it too low there can be millions of sequences.  So it need to be adjusted for each dataset
2. This is not a problem.  You need to create a dictionary containing X= A->B. Then during the prediction of a sequence when you read X you look up in the dictionary to see that it is A->B. Thus the prediction process will give the same result. 
   By the way,  if you make a good Python version of CPT or CPT+,  I think that it would be great if you publish your code as open source. 

Sent from Yahoo Mail on Android

On Wed, Aug 10, 2016 at 17:30, Hassanbenlebsirnotifications@github.com wrote:
Hi,
Brilliant work your article is really compréhensive and you made everything so clear.

i'm working on a subject and i'm coding CPT in python.

subject describtion
my problem is based on the past for instance this is my past A-->B
I want to know the out come for this past.
so i look at my tree and i search for all sequences with the same past and see the outcome it.
example: A-->B-->C so for this case C is the outcome for this past (that means C+=1) and i do this for the whole tree
so i can suggest the outcome for the past A-->B .

until now i can do this with CPT and i want to improve it and go to CPT+ using FSC.
I followed your steps i calculate the frequency of every subsequence and i fixed a minsub

questions
My first question:
How to fix this minsub ?
second question FSC:
For this this question i'll give an example first
ex:we have this path A-->B-->C-->D and we know that the B-->C is frequent so i replace B-->C with X : A-->X-->D.
so now when i want to calculate the outcome for my past for example A-->B
the subsequence A-->B it will not be found in this case because B is a part of X so is there away to avoid this?

I'm looking forward to your reply

—
You are receiving this because you are subscribed to this thread.
Reply to this email directly, view it on GitHub, or mute the thread.