Supervised, Unsupervised & Reinforcement Learning

Organisations are exploring and generating real returns on their investments from Machine Learning, namely, Supervised Learning, Unsupervised Learning and Reinforcement Learning. It’s a science that’s not new, but one that has gained fresh momentum due to unprecedented access to compute and storage on cloud platforms enabling organisations to quickly and automatically produce models that can analyse bigger, more complex data and deliver faster, more accurate results at scale.

Organisations are exposing underlying data patterns that were previously undiscovered, predicting behaviours and outcomes where there are many complex predictors, and training models to forecast scenarios before implementing expensive solutions, products and services. These models also unlock and identify new opportunities and help avoid previously unknown risks.

Request Consultant

Supervised

ML Process & Requirements

  • Gather and centralise the data required for machine learning.
  • Understand data statistics and outliers, hold workshops with SMEs to understand nuances in the data and make decisions.
  • Prepare data including targets, labels, features and statistics.
  • Test and choose initial machine learning models/algorithms.
  • Train the model using a training data set (supervised learning).
  • Evaluate, tune, reevaluate and retune the model as necessary (iterate) to improve prediction rates and accuracy.
  • Make recommendations and continuously improve predictions and understanding of emerging trends that might appear.

Supervised

Supervised Learning

 

The learning algorithm/model receives a labelled training data set, which is used to retrospectively identify predictors to create a predictive model. Some examples include:

  • Segmentation, customer value, churn, purchase behaviour, promotions, advertising, chatbots.
  • Supply and demand forecasting and optimisation.
  • Predicting values and categories based on predictors.
  • Video, audio and image classification, object detection, face recognition and event detection using Deep Learning.

Unsupervised Learning

The goal is to explore the data and find some structure (including clusters and associations) without receiving any training data. Some examples include:

  • Understanding underlying clusters and associations in social networks, customers and market segmentation.
  • Association mining for basket analysis (understanding goods/services commonly purchased together).
  • Identifying anomalies in datasets including fraud, faults, unusual behaviour and human error.

Reinforcement Learning

The learning model discovers through trial and error which actions yield the greatest rewards. Some examples include:

  • Understand/optimise complex traffic and process flow.
  • Optimise resource allocation (dollars, time, effort, etc.)
  • Machine, equipment and plant tuning & optimisation.
Related Content
View All