AutoML Pipeline

Machine learning model development is associated with tedious and repetitive steps in a pipeline. You can automate the steps required in the model development using Automated Machine Learning, commonly known as AutoML.

Building an ML model includes the steps like data cleaning, data pre-processing, feature engineering, model selection, training, hyperparameter tuning and, evaluation and most of them are repetitive in nature. This entire sequence of steps are automated using AutoML by generating automated pipelines with best possible configuration. The result of this automated pipeline is the finalised version of a model with the best possible performance evaluated by dedicated evaluation metrics depending on the type of the model.

Key components of an AutoML pipeline include:

• Data Preprocessing: Optimizes input data quality by performing data cleanup and transformation operations, ensuring the data is ready for model training.
• Feature Engineering: Automatically selects and removes features or creates new ones as required, thereby improving data quality and resulting in better model performance.
• Model Selection: Selects the most suitable algorithm from an available list that gives the best performance of a model, ensuring the chosen model is well-suited for the data addressing the specific use case.
• Hyperparameter Tuning: Determines the optimal values of hyperparameters by optimizing the loss incurred during training of the chosen models, resulting in enhanced model performance and robustness.
• Model Evaluation: AutoML pipeline generates multiple models during training, which are evaluated using predefined evaluation metrics. The best-performing model will then be selected.

AutoML aims to make machine learning accessible to novice users by automating these complex tasks, also, enabling experts to create high-quality models efficiently in less time. This allows users to concentrate on addressing real-world problems, understanding the results, and applying the insights to solve them.

Steps to create AutoML pipeline in QuickML

1. Navigate to the bottom-left of the page, click the +Create button, and select Pipeline.
2. The Create Pipeline page will appear. Fill in the details, then check the Create an Auto-generated pipeline using AutoML checkbox.
   
3. Click Create Pipeline. This will create an automated pipeline with the necessary configuration, as shown below:
   

The best performing model will be finalized and generated with all the evaluation metrics, as shown below.