Complex Event Processing/Event Stream Processing

From my understanding of CEP, there seems little value as rosy mentions.

It's a fancy way of saying:

Dynamic Correlation Analysis
Intermarket Behavorial Pattern Analysis

These have been around the days of Bayesian Processing and Machine Learning. They've been around for over 10 years... and has a far higher significance for trading.

In true forms, Algorithmic Trading is a fancy word for order-routing technology to reduce liquidity risk. High Frequency trading is another word for short term arbitrage and/or scalping models.

For any model, you need fast and accurate data. If you were to filter the query, the only reason to do so is to detect back-fill or any bad data. You do not need anything influencing the significance of any data. That's what the arb model does... ask... Why do you have an arb. model to start off with... duh...

You're a trader, don't believe the hype.

Or... am I missing something?
 
Yeah, that was really my point. I dont beleive the hype.

We already run our models in a custom environment, and were extremely happy with it in terms of flexibility and speed.

Just interested in other punters veiws on the subject, and input from anyone who has implemented the technology, and found that it either exceeded or fell well short of their expectations.

No answers from men in shiny suits please.
 
Quote from zzt:

Ok. For example, would the query language enable me to extract eigenvalues from a correlation matrix ?

with esper you can write your own user defined functions. the query language is like a databases's sql, so once you want to do more than the basics you need to add to it.
 
Quote from TSGannGalt:

From my understanding of CEP, there seems little value as rosy mentions.

It's a fancy way of saying:

Dynamic Correlation Analysis
Intermarket Behavorial Pattern Analysis

Actually, no. It's another way of saying: In-Memory Database.

CEP/ESP aren't the actual algos, they're the framework that algos get written to/with. I doubt it's anything more complex than super optimized arrays, with accessor functions that take sql-like queries.

Probaby a better way of describing these are: Yet Another SellSide Gimmick. But, I guess if you don't have the knowledge to roll your own, and have uber cash, you just purchase a solution.
 
Quote from onelot:

Actually, no. It's another way of saying: In-Memory Database.

CEP/ESP aren't the actual algos, they're the framework that algos get written to/with. I doubt it's anything more complex than super optimized arrays, with accessor functions that take sql-like queries.

Probaby a better way of describing these are: Yet Another SellSide Gimmick. But, I guess if you don't have the knowledge to roll your own, and have uber cash, you just purchase a solution.

http://en.wikipedia.org/wiki/Complex_Event_Processing

Complex Event Processing, or CEP, is a technology for building and managing event-driven information systems. CEP is primarily an event processing concept that deals with the task of processing multiple events from an event cloud with the goal of identifying the meaningful events within the event cloud. CEP employs techniques such as detection of complex patterns of many events, event correlation and abstraction, event hierarchies, and relationships between events such as causality, membership, and timing, and event-driven processes.


Are we on the same subject? Because I'm definitely not catching on...

:confused: :confused: :confused:

Added on edit...

I've read a few on this topic so far. I still don't see why it's a big deal. I guess it's wanna-be-Algorithmic Trading-for-dummies...

WHEN Person. Gender EQUALS “man” AND Person. Clothes EQUALS “tuxedo”
FOLLOWED-BY
Person. Clothes EQUALS “gown” AND
Church_Bell OR
Rice_Flying
WITHIN 2 hours
ACTION Wedding


Sounds like a language for people who can't write a simple conditional statement.

Or am I STILL missing something???

:D :D :D
 
I don't understand why everyone in this thread is so defensive. Esper is free, there aren't any "gimmicks". Here is one example of how I am using it (maybe this will help someone):

queryText.append("insert into harmonics_window_"+side+"\n");
queryText.append("select\n");
queryText.append((y+1) + " as index,\n");
queryText.append("count(*) as counter,\n");
queryText.append("marketMaker as marketMaker,\n");
queryText.append("coalesce(sum(price * priceLevel * size * weight), 0) as harmonic_val\n");
queryText.append("from LevelII"+side+"Event.std:groupby('marketMaker').win:time(" + ((y+1)*intervalSecs) + " seconds)\n");
queryText.append("group by marketMaker\n");
queryText.append("output every "+(55*intervalSecs)+" seconds\n");

I considered rolling out my own ESP, but this was just too easy. Esper handles all of the in-memory caching and keeps the sliding window in seconds or ticks. The greatest advantage is the ability to match events by "marketMaker" and then perform aggregation over the sliding window, grouped by each market maker.
 
Quote from snaggydog:

The above is completely false and misleading. Event stream processing (ESP) is "in-memory databases". CEP is more more complex.

See, for example:

http://tibcoblogs.com/cep

Object oriented, finite state machine, blah, blah - but what does it actually do ?

I'm not being hostile at all, but to be realistic he could be talking about just about anything.
 
Back
Top