question

23174178 avatar image
0 Votes"
23174178 asked shashishailaj commented

Error - WebApp Implementation with Principal Component Analysis(PCA)

After applying Principal Component Analysis PCA to my data set in order to achieve better model accuracy. The 13 features dimensions, I am reducing it to 10 features using PCA. Everything is fine till here.

After implementing the model in WebApp, it is building & seems fine in the studio.

In the testing phase of model prediction, Instead of displaying 10 features as an input, the UI system is showing the original features which is 13 & the output is showing 10 featu70351-01.pdfres which does not have any feature names for the newly generated features which are 10. And also prediction is not working at all after executing it.\

Attached are the screenshots, Please refer.



azure-machine-learning-studio-classic
01.pdf (54.5 KiB)
5 |1600 characters needed characters left characters exceeded

Up to 10 attachments (including images) can be used with a maximum of 3.0 MiB each and 30.0 MiB total.

1 Answer

romungi-MSFT avatar image
0 Votes"
romungi-MSFT answered 23174178 commented

@RakshitSidd-7739 After updating and running the training experiment did you try to update the prediction experiment.

70512-image.png

After the prediction experiment is updated you can update the web service and check if it shows up correctly.


image.png (86.1 KiB)
· 7
5 |1600 characters needed characters left characters exceeded

Up to 10 attachments (including images) can be used with a maximum of 3.0 MiB each and 30.0 MiB total.

Hello @romungi-MSFT , I did try. It is not successful.
The problem here I am facing is, After dimensionality reduction technique which is PCA. In the WebApp testing phase, Input feature should have been 10 & so as output. But the input feature is showing as 13 feature & the output is 10 feature, which is not matching and does not make sense.

Can anyone in the community here please explain with steps. As I am new using this Azure ML Studio.

0 Votes 0 ·

@RakshitSidd-7739 I think based on the configuration of your PCA module, it outputs a reduced set of columns that you can use in creating a model. But the input to the web service will still be the same i.e 13 as the module itself is trained to take the selected columns as input and apply normalization on this data. Also, as per the note in the configuration of this module:

The algorithm functions optimally when the number of reduced dimensions is much smaller than the original dimensions.



0 Votes 0 ·
image.png (13.8 KiB)

Dear @romungi-MSFT , Yes I get this concept. But what should I do If I have to achieve those 10 feature as input & output after building WebApp with Principal Component Analysis technique ?

I feel, I am missing some steps during the process of prediction model even before WebApp build.

0 Votes 0 ·
Show more comments