spark-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Jörn Franke <>
Subject Re: A naive ML question
Date Sun, 29 Apr 2018 12:44:48 GMT
The transactions probably describe from which counterparty assets are transferred to another
counterparty at the different stages of the transaction. You could use graphx for that if
the algorithms there are suitable for your needs.
Still trying to understand what you mean evolve over time? Eg a counterparty has cancelled
a lot of transactions or sth like this? Normally it looks like you have a rather straight
forward state machine for your transactions. 

> On 29. Apr 2018, at 12:18, kant kodali <> wrote:
> Hi Nick,
> Thanks for that idea!! Just to be more clear. The problem I am trying to solve is that
when a bunch of financial transactional data is thrown at me I am trying to identify all possible
relationships and lineage among them without explicitly specifying what the relationships
are among transactions.
>> On Sun, Apr 29, 2018 at 2:22 AM, Nick Pentreath <>
>> One potential approach could be to construct a transition matrix showing the probability
of moving from each state to another state. This can be visualized with a “heat map” encoding
(I think matshow in numpy/matplotlib does this).
>>> On Sat, 28 Apr 2018 at 21:34, kant kodali <> wrote:
>>> Hi,
>>> I mean a transaction goes typically goes through different states like STARTED,
>>> Thanks,
>>> kant
>>>> On Sat, Apr 28, 2018 at 4:11 AM, Jörn Franke <>
>>>> What do you mean by “how it evolved over time” ? A transaction describes
basically an action at a certain point of time. Do you mean how a financial product evolved
over time given a set of a transactions?
>>>> > On 28. Apr 2018, at 12:46, kant kodali <> wrote:
>>>> > 
>>>> > Hi All,
>>>> > 
>>>> > I have a bunch of financial transactional data and I was wondering if
there is any ML model that can give me a graph structure for this data? other words, show
how a transaction had evolved over time? 
>>>> > 
>>>> > Any suggestions or references would help.
>>>> > 
>>>> > Thanks!
>>>> > 

View raw message