Automated Machine Learning (AutoML) in Azure

Automated machine learning, known as automated ML or AutoML, streamlines the tasks involved in developing machine learning models by automating repetitive tasks. Azure Machine Learning provides the Python SDK, allowing you to leverage the power of AutoML.

Applications of AutoML

It simplifies the machine learning model development process, allowing users to implement ML solutions without extensive programming knowledge.

Classification

• Identifies data points into categorical labels

• Precision is the ratio of correctly predicted positive observations to the total predicted positives. It measures a model’s accuracy in making positive predictions.

• Confidence refers to the probability assigned to a model’s prediction, indicating its likely accuracy.

• Used in scenarios like fraud detection, object detection, or handwriting recognition.

• Utilize Azure components such as:

  • • Azure Machine Learning – training and deploying machine learning models.

    • Responsible AI Toolbox – identify contributing factors and detect bias in the model.

    • Azure Synapse Analytics – utilized for efficient data processing.

Regression

• Analyzes relationships between continuous variables

• Predicts numerical outcomes

• Mean Absolute Error (MAE) is a metric that calculates the average of absolute differences between predicted and actual values, representing the prediction error.

• It is used in scenarios like predicting sales figures, estimating housing prices, or forecasting stock market trends.

Clustering

• It is used to group similar data points together based on their similarities without any predefined categories or labels.

Time-series forecasting

• Predicts future values based on historical patterns in sequential data
• Used in scenarios like demand forecasting, inventory management, or sales forecasting.

Computer vision

• For use cases that primarily requires input from visual data

• Aims to extract meaningful information from images or videos

• A computer vision approach would involve training a model on a dataset, that are used in scenarios like interpreting handwriting, differentiating digits, pattern recognition, or virtual objects.

• Automated ML for computer vision supports:

• • • Multi-class image classification

    • Multi-label image classification

    • Object detection

    • Instance segmentation

• Azure AI Vision, a unified service for innovative computer vision capabilities. It is used by apps to analyze images, face detection, text extraction with OCR and facial recognition.

Natural language processing (NLP)

• For use cases that primarily requires input from text or speech data

• Aims to analyze, understand, and generate human language

• It focuses on understanding and processing human language, which are used in scenarios like text/sentiment analysis, language translation, review insights, or automated responses.

Azure Machine Learning Models

Azure Machine Learning empowers data scientists to improve model performance and accurate predictions in a customizable manner.

Feature Engineering

• A practice of transforming raw data into a format that ML algorithms can understand and make better predictions.
• Scaling and normalization techniques are used in Azure Machine Learning to facilitate Featurization.
• Incorporates steps like feature normalization, handling missing data, converting tests to numeric, etc.
• Customize featurization
  • • Additional feature engineering techniques available: encoding and transforms
    • Enable Automatic featurization in Azure Machine Learning Studio or specify featurization in Python SDK.

Ensemble Models

• Ensemble Models is a practice of combining multiple models to improve overall predictive performance.

• Voting and stacking ensemble methods in Azure Machine Learning.

  • • Voting – takes the average of multiple model predictions

    • Stacking – default meta-model that trains a meta-model to learn from the predictions of different models

• Caruana ensemble selection algorithm for effective model inclusion.

  • • create an ensemble by selecting the best models and by adding new models that enhance the overall score

References:

What is automated ML? AutoML – Azure Machine Learning

Featurization with automated machine learning – Azure Machine Learning

Predict student attrition – Azure Example Scenarios

What is Azure AI Vision? – Azure AI services