Question on Value of Probability of value 1 or 2 in Neural Network Viewer

Hi, all,

I am confused about the value of Probability of Value 1 or 2 (on a particular attribute value) in Neural Network viewer. E.g. the value of Probability of value 1 is actually very low (the same to the value of Probability of value 2), but why the bar which shows the strength of the probability of these two values are still so strong even stronger than other values of probability of value 1 or 2 based on other attribute values which have a much higher probability of value 1 or 2?

And how does the algorithm calculate the Probability of attribute value in nerual network by the way?

Hope my question is clear.

I am looking forward to hearing from you shortly and thanks a lot in advance.

With best regards,

Yours sincerely,

The scores represent the predicted probability as influenced by a single attribute value.

Here is how the scores are computed:

Assume that the model predicts Bike Buyer (= TRUE or FALSE), based on Home Owner, Commute Distance etc.

A stored procedure creates a "virtual" input, containing, for each row, a single attribute (everything else is missing). Example:

Row 1: (empty) -- all attributes are missing

Row 2: Home Owner = TRUE, everything else is empty (missing)

Row 3: Home Owner = FALSE, everything else is empty (missing)

Row 4: Commute Distance= 0-2 miles, everything else is empty (missing)

Row 5: Commute Distance= 2-4 miles, everything else is empty (missing)

and so on.

For each row in this "virtual" input, the procedure computes PredictProbability(BikeBuyer, 'TRUE') and PredictProbability(BikeBuyer, 'FALSE') as well as Predict(BikeBuyer)

All the results that favor TRUE are then normalized in the 0-100 space

Same for all the results that favor FALSE

So, practically, the Probability of Value 1(or 2) represents a normalized view of the prediction probability for value 1 generated by that input.

Because of the normalization, inputs that predict strongly a value may still have low scores (simply because all other inputs are stronger)

Hope this helps

|||

Hi,

Thanks for that.

With best regards,

sql

Question on Links coefficiency of dependency network

Hi, guys,

Thanks for your kind attention.

Just wonder is there any way to get the link value on dependency network? Yes, I know we can move along the slider to get the link strength between different attributes. However, it will be more convinced that if we can get the value of the links and say we have a standard for it to judge if the link is truly strong enough?

Thanks in advance for your advices.

And I am very much looking forward to hearing from you.

With best regards,

Yours sincerely,

You could use the Profiler to see the calls executed against the server when the dep net viewer is populated.

If you execute the stored procedure from, say, SSMS, you will get back the whole serialized graph, with the weight associated with each edge

|||Data Mining Add-in for Visio can show edge strength (there is a button for it in the Data Mining toolbar).|||

Hi, thanks for the advices.

Best regards,

Yours sincerely,