LAE

Note: This was originally posted by an inactive account. Content was preserved by moving under an admin account.

Originally posted by: michaelslowey

Hi all,

Big ask....

Has anyone used LAE to produce a trend for a set of data. eg. I have a set of values between 0 and 16,000 and I want to output that to a graph, but I also want to create a second set of values with a calculated polynomial trend value to show against the actual in the graph....

Cheers
Mick

Note: This was originally posted by an inactive account. Content was preserved by moving under an admin account.

Originally posted by: timonk

Mick,

May I clarify; Are you looking to produce a dataset(s) g(X) from your column X, and then display an actual chart of the results for comparison from within LAE?

Regards
Timon

Note: This was originally posted by an inactive account. Content was preserved by moving under an admin account.

Originally posted by: michaelslowey

I am looking to produce a second column (Trend Value) from the first column.

Since the trend value will depend on more than one value in the first column I'll have to use an agg but really what I am looking for is anyone who has used a complex algorithm to produce a trend from a set of data...The charting is secondary and I can deal with that later.

Cheers
Mick

Note: This was originally posted by an inactive account. Content was preserved by moving under an admin account.

Originally posted by: vmilligan

Mick

I am not sure exactly what you are looking to do and why. It seems that you have data that is not presenting itself into clear enough patterns - so you are looking for different trend visuals to see if they may expose them? What I may suggest is to break the data into smaller indexed chunks to see if smaller patterns emerge, or check that the data is "supposed" to say, and find variations to the resultant profile.

Note: This was originally posted by an inactive account. Content was preserved by moving under an admin account.

Originally posted by: michaelslowey

Victor, finally figured this one out.


This node takes a set of data and applies a linear regression algorithm to that data set, comes up with the m and b values and applies that to the data set to determine the trend of the data over a period of time.

You may be able to see the trend by eye if you plot the data values, but if you have 100,000 subs each with a different trend you'd be unable to eyeball them all so i wanted a flag or a value that we could divine in a graph to segment the customer as a growing or diminishing customer.


I'm using this in a churn graph to determine over a period of time if a customer is reducing or increasing the number of calls made. A negative m value is decreasing and a positive m is
increasing..

Mick
Attachments:
linear_regression.zip