How to view 5 outputs using Microsoft Neural Network using SSAS

Question

Hello there,

 I have a doubt using microsoft neural networks.

Here we have this network output attribute which supports 2 values or states which is either a Yes or No.

Is there any way I can use the same neural networks which can support 5 outputs and also view it using the viewer.

PS: I have tried changing the network parameters where I changed the maximum output parameters as 5. Yet I am only able to view 2 values.

Any help would be appreciated

Regards

  

Answer 1

The network is built for all 5 output states, and the predictions will function correctly. The visualization is designed to emphasize patterns that discriminate between any 2 states. I would suggest the following: - change Value 1 to the first value in your 5 states - Change Value 2 to AllOtherStates - Save the report in Excel or something similar - repeat for all distinct 5 states (in Value 1) The result will show you the strong predictors for each of the 5 distinct states

---

bogdan crivat / http://www.bogdancrivat.net/dm

Answer 2

Dear bogdan,

 Thank you so much for your response. I shall try that out and post the status here. 

Regards

 

Answer 3

Dear Sir,

  I tried the method you suggested. But The value column doesn't seem to support AllOther States. The value column had a "missing" state and then all the other 5 states.

The attribute column had an all state.

I also checked the export option to exdport the file to excel. But SSAS exports files to a note pad in xml format.

Any more valuable suggestions would be nice

regards

 

Answer 4

Hi vaishaks,

According to the Neural Network Algorithm Technical Reference,

Output neurons represent predictable attribute values for the data mining model. For discrete input attributes, an output neuron typically represents a single predicted state for a predictable attribute, including missing values. For example, a binary predictable attribute produces one output node that describes a missing or existing state, to indicate whether a value exists for that attribute. A Boolean column that is used as a predictable attribute generates three output neurons: one neuron for a true value, one neuron for a false value, and one neuron for a missing or existing state. A discrete predictable attribute that has more than two states generates one output neuron for each state, and one output neuron for a missing or existing state. Continuous predictable columns generate two output neurons: one neuron for a missing or existing state, and one neuron for the value of the continuous column itself. If more than 500 output neurons are generated by reviewing the set of predictable columns, Analysis Services generates a new network in the mining model to represent the additional output neurons.

Could you give us an example to present the real case in your project for us to better understand your requirement? Maybe there are some workarounds for it.

Regards,
Jerry

Answer 5

The model you created will correctly predict all 5 states.

I think you are confused because Microsft Neural Network viewer shows patterns that discriminate between any 2 states. Unfortunetly it does not even allow to discriminate between 1 state and all compliments of that state (like Bogdan suggested).

I think only small percentage of people who use Microsoft Neural Network algorithm find some useful information in the viewer. Most people just apply the algorithm without understanding what exactly it is doing. (Many people also choose to use Decision Trees or other algorithm even if it is not as accurate as Neural Network because they can understand the patterns in the algorithm).

---

Tatyana Yakushev [PredixionSoftware.com]