How to Create a Custom Machine Learning Model?

2 mins read - Updated on Sep 26, 2022

Analyzing reviews, feedback, and conversations of consumers is a never-ending process for marketing pros, founders, researchers, and any dataholic. If you’re working for an enterprise-level brand and constantly tracking conversations, ready-made ML models at Kimola Gallery might not solve your problem, so it’s best to create a custom ML model. With a custom ML model, you can decide the labels that your content will be classified and keep classifying the upcoming data. 

So let’s dive in to see how you can create a custom ML model. 

First, Sign in and Go to Kimola Cognitive home page.

If you’re already a member of Kimola Cognitive, the first thing you should do is to sign in and go to your home page of Kimola Cognitive. On your home page, click the plus icon as shown in this screenshot:

Screen Shot 2022-05-10 at 18.12.38   

Onboarding Screen

Now, the onboarding screen should appear. This screen is a walkthrough to create your new ML model.

Screen Shot 2022-05-10 at 18.12.54

As also stated in the onboarding screen, you will need to upload a dataset to train your ML model. 

What is a dataset for training?

A custom ML model is mostly created to analyze thousands of conversations with defined labels. Your training set should contain texts, as Kimola Cognitive only analyzes text data. Here is an example of a dataset to upload for training:

Screen Shot 2022-05-10 at 18.13.02

A Quick Guide to Create The Best Dataset for Training:

    • The text content must be about a topic, a brand or anything that can be classified
    • The data should be classified with a minimum of two and a maximum of eight labels
    • Each label should have a minimum of 500 rows of text.
    • Content should be distributed equally on each label.
    • The data should be cleaned: A clean dataset means that links, nouns, numbers, and names should be deleted from each row. Here is a quick guide to deleting those with Google Sheets automatically

Finalize Onboarding Steps (You can also watch the video below!)

  • Upload your dataset: Drag and drop your dataset on the screen. 
  • A preview of your dataset should appear. Choose the text column and label column.
  • Give a name to your custom ML model. 
  • Choose the language of your training set. You can only choose 1 language, so your training dataset content should have a single language.
  • Click Complete!
  • An e-mail will be sent to you when our monster is trained and your ML model is ready. 









Was this article helpful?

Tell us about your thoughts and experiences regarding the article.