How to leverage Predictive Intelligence on imperfect data?

[Suggy]

Customer is looking to implement Predictive intelligence but the exisitng data is improper and not clean.

Ex - not all the incidents are properly categorized. For few incidents, category itself is not set. etc.

How can we use Predictive intelligence/Task intelligence in this case?

[abirakundu23]

Hi @Suggy ,
You can create Background script to create dummy data in your instance to implement Predictive Intelligence.
Basically you have to trend the data so that as per configuration minimum data set range should be available for trained the model as per documentation.

Please mark helpful and correct answer if it's worthy for you.

[Suggy]

@abirakundu23  my question is different 🙂

[abirakundu23]

Hi @Suggy,
In that case you can generate report for those  data and update those record by fix script or back ground script.
Focus on recent high volume data for train the data.

Train Predictive Models based on the Clean Subset of data.

You can build model accordingly.

Categorization model: Predict category/subcategory

Assignment model: Predict assignment group

Apart from that what are you looking exactly ? Please mention clearly so that I can support to you.
Please mark helpful and correct answer if it's worthy for you.

[Suggy]

Hi @abirakundu23  Customer would not allow to touch the closed records as its not best practice. if they allow as well, we have close to 3 lac incidents. So correcting them would be nightmare.

We have informed that with bad quality data or incomplete data, they cannot use PI/TI.

but still wanted to check once in community if anyone any thoughts on this 🙂