Train ML models

Training Machine Learning Models on Octai

We believe that building machine learning models should be accessible to everyone, regardless of their coding or technical skills. That's why we've created Octai, to enable you to build and deploy models quickly and easily, with no coding required. Simply select your training data and target (what you want to predict), and the platform handles the rest for you.

Octai offers fast experimentation, enabling you to build and test multiple models in a fraction of the time it would take to do manually. This helps you to iterate quickly, experiment with different approaches, track your setups, and optimize your models for maximum accuracy.

Training machine learning (ML) models in the Octai platform involves a series of steps that simplify and automate the process for users. Here's a general outline of how to train ML models using Octai:

  1. You can visit the "Model" module for in-depth documentation about the whole modeling and experiment tracking.

You can select the data and label it on the simplest level, then click "Train".

Auto training handles these processes by optimizing the configuration based on your data.

The platform will iteratively train and evaluate the models on the training and validation datasets, respectively, optimizing their performance based on the specified evaluation metric(s).

Auto Train

Auto Train

  1. Or you can select manually. Specify the target variable (the variable you want to predict), the evaluation metric(s) used to measure model performance, and any other necessary settings or constraints, such as the type of problem (regression, classification, etc.), or specific algorithms to include/exclude.
  2. Divide the data into training and testing (or validation) sets. Octai usually handles this step automatically, ensuring that the models are evaluated on unseen data to prevent overfitting.


  1. Review the results to see which model performed the best. Octai typically provides a leaderboard or summary of the trained models, along with their performance scores and other relevant information.


  1. Evaluate the best model: Further, evaluate the best-performing model on a separate test dataset (if available) to assess its ability to generalize to unseen data.
  2. Deploy the model: If you're satisfied with the performance of the best model, deploy it to a production environment, where it can be integrated into applications or services to make predictions or decisions based on new, unseen data.