Results of DMX queries in MovieClick association rules model

Question

Hi,

I'm trying to practice DMX on association rules. I have deceloped the MovieAssociation model on MovieClick database as per "data mining with SQL Server 2008" book. I tried the DMX provided by the book as well as the DMX simillar to what has been recommended in a posting here. I get no errors but I get odd results for any of the queries below. It sounds like it's giving just the 5 top purchased movie regardless of what I query. Am I missing something here?

Thanks




// (here Outcome recommendation is same for all customers)

SELECT T.CustomerID, Predict(Movies, 5) AS Recommendation
FROM MovieAssociation
NATURAL PREDICTION JOIN
  OPENQUERY([Chapter 11],
  'SELECT CustomerID, Gender, [Marital Status] FROM Customers') AS T


//Result: (for all customrt IDs)

//Star wars
//Matrix, The
//Lord of the Rings: The Fellowship of the Ring, The
//A beautiful mind
//Star Wars Episode V: Empire Strikes Back

 

//(here I get same result below for queries regardless of gender type I mention)

SELECT Predict(Movies, 5) FROM MovieAssociation PREDICTION JOIN

( SELECT 'Female' AS gender) AS T

ON MovieAssociation.gender = T.gender

//Result: (for both male or female options)

//Star wars
//Matrix, The
//Lord of the Rings: The Fellowship of the Ring, The
//A beautiful mind
//Star Wars Episode V: Empire Strikes Back

SELECT Predict(Movies, 5) FROM MovieAssociation

PREDICTION JOIN

(select 'male' as gender, (SELECT 'star wars' AS movie union select  'big' AS Movie) as Movies)  as T

ON MovieAssociation.gender = T.gender

and MovieAssociation.Movies.Movie= T.Movies.movie

//Result: (for both male or female options)

//Star Wars Episode V: Empire Strikes Back
//Star Wars Episode VI: Return of the Jedi
//Star wars
//Matrix, The
//Lord of the Rings: The Fellowship of the Ring, The

Answer 1

Hi Jeff

What results do you get for 

SELECT Predict(Movies, 5) FROM MovieAssociation

I am going to guess the exact same as you have currently.  The above query should give you most probable 5 movies selected in the model

What your query is telling you I think is "look I couldn't really find anything interesting for you with the data you gave me so i am going to give you the model's most likely movies instead"

Allan