Can we train and deploy CUSTOM machine learning model in Servicenow?

[Sharique Azim]

Hi All,

Recently i found some posts where people have deployed custom logic and machine learning(ML) models in servicenow using AWS cloud services(AWS lambda) which is implemented on REST API architecture ServiceNow - Machine Learning. I have also gone through the post where Tensorflow.js has been integrated with servicenow to predict results in browser. TensorFlow.JS (Prediction)+ ServiceNow . Thanks to both of you for your work.

Now  my question is:

1.Can we deploy and train on our own custom ML models on custom table, without using any REST API to any cloud service?

2.If yes, then can we use the same  to  save the predictions in some table and call it from the browser/client using some UI script or script include?

3.If we cannot train on our  custom models ,then how can we load/use the trained data  set into servicenow, such as to call from UI script or script includes.

Please note, we would be using this in either UI page or service portal.

Any help would be highly appreciated.

Regards,

Shariq

[Chuck Tomasi]

You basically have two options that I can see (perhaps someone else has others I haven't thought of.)

1. Use a REST API to a third party service like AWS to train and retrieve predictions
2. Use ServiceNow's Agent Intelligence product.

[Sharique Azim]

hi chuck,

So accodring to you, do you mean there is no way to implement custom models in servicenow.without hitting 3rd party Cloud API?

[Chuck Tomasi]

See option 2 - you can create models against your tables with Agent Intelligence provided you have enough records (at least 30K) to train the model.

[Chuck Tomasi]

I stand corrected. I have corrected the previous post. The answer is not MetricBase. That's a time series database best used for gathering massive non-related information like IoT data.

Agent Intelligence is the machine learning based product that you create a model via training on existing data. Sorry for the confusion.