Predictive Analytics and Machine Learning for Business

Predictive analytics has revolutionised business across a range of industry verticals.

Predictive analytics has revolutionised business across a range of industry verticals, including banking, insurance, healthcare, retail, telecom, media and e-commerce. Despite the returns on predictive analytics being well documented, companies often proceed cautiously when presented with new predictive analytics initiatives. And with good reason – poorly built predictive models are costly and useless to the organisation. With a strong emphasis on the proven machine learning models used by leading companies such as Google, Walmart and AIG, this course provides a solid grounding in machine learning methods including classification, random forests and decision trees.

This course is taught in R and Azure.

Who Should Attend?

This course builds core skills for those who require a grounding in predictive analytics. Candidates who will benefit from the course include:

  • Graduate hires who are on track toward a career in analytics or data science
  • Managers and executives who wish to use data to predict business unknowns
  • Candidates already working in analyst positions, including (but not limited to) positions involved with data analytics, digital and marketing analytics, customer and market analysis
  • Any other business professional who would like to make decisions on the basis of data

Prerequisites

It is recommended that students have completed Introduction to Data Science and Analytics, or have an undergraduate degree in statistics or mathematics.

Laptop Required Specs

Intel i3 processor, 4GB RAM

Either Mac or Windows operating system

Software Requirements

Excel 2010 / 2013 / 2016

R or RStudio Latest version

A free trial or paid subscription to Microsoft Azure ML Studio

Course Objectives

  • Provide participants with a full understanding of multiple linear regression, logistic regression, market basket analysis and decision trees
  • Provide participants with a full understanding of model diagnostics and underlying assumptions for a range of predictive models that are appropriate in the business context
  • Provide participants with a framework for selecting an appropriate predictive model depending on the task at hand
  • Provide participants with a qualitative and quantitative methodology for assessing the performance of predictive models
  • Provide participants with a firm grounding in the business case for Market Basket Analysis and the mathematics underlying this technique
  • Provide participants with guidelines on how to present predictive analytics results in the corporate setting (taught via simulated boardroom presentations in class)

Upon successful completion of this course, participants will be able to:

  • Build and use predictive models to inform decision making in the face of uncertainty
  • Build robust predictive models that predict sales, revenues, employee attrition, customer churn or other events
  • Know when to apply a classification algorithm instead of a continuous prediction algorithm
  • Understand the limitations and benefits of predictive models and how to communicate these limitations and benefits to senior decision makers
  • Separate esoteric and academic predictive models from those that are proven to be robust in the business setting
  • Understand how to calculate the ROA (return on analytics) for predictive analytics projects
  • Recognise business opportunities where the investment in predictive analytics is justified in the context of project risk and expected benefits
  • Understand how to manage predictive analytics project risk proactively whilst preventing time and cost overruns

Here’s an outline of what will be covered during this 3-day workshop

Day 1

+

Day 1

I. Introductions (9:00am – 9:30am)

II. Dimensionality, Parsimony, Testing Accuracy (9:30am – 10:00am)

  • The curse of dimensionality
  • The principle of parsimony
  • Testing model accuracy
  • John Elder’s Target Shuffling
  • Lift charts
  • Bootstrap sampling

III. Q&A / Break (10:15am – 10:30am)

IV. Shrinkage – More Than What Happens in the Pool (10:30am – 11:00am)

  • How shrinkage methods depart from traditional statistical methods
  • Ridge regression
  • The LASSO method
  • How does the LASSO method help perform variable selection?
  • Sparsity

V. Q&A Break (11:00am – 11:15am)

VI. Workshop: Team Activity – Let’s compare LASSO and ridge regression (11:15pm – 12:00nn)

VII. Lunch (12:00nn – 1.00pm)

VIII. Workshop: Team Activity (cont.)- Let’s compare LASSO and ridge regression (1:00pm – 1:30pm)

IX. Cross Validation, Bagging and Ensembling (1.30pm – 2:15pm)

  • Bootstrap aggregation
  • K-fold cross validation
  • Model ensembling
  • Choosing weights for ensemble models

X. Q&A / Break (2:15pm – 2:30pm) 

XI. Workshop: Let’s bag, ensemble and cross validate! (2:30pm – 4:45pm)

XII. Workshop Feedback, Presentation from Winning Model and Day 1 Wrap Up  (4:45pm – 5:00pm)

Day 2

+

I. Artificial Neural Networks (9:00am – 10:00am)

  • A gentle introduction to ANNs using colors
  • What is deep learning?
  • What is forward and back propagation?
  • How many hidden layers should we use?
  • ANN and linear regression smack down in Azure ML Studio

II. Q&A / Break (10:00am – 10:15am)

III.  Workshop: Team Activity – Let’s build and tune neural nets (10:15am – 12:00nn)

IV. LUNCH (12:00nn – 1:00pm)

V. Predictive Analytics in Practice – Managing Analytics Projects and Teams (1:00pm – 1:45pm)

  • Where should the analytics team be situated in the corporate structure? Research findings.
  • Managing stakeholder expectations in analytics projects
  • The importance of having analytics champions
  • Project management for analytics projects – how does it differ from regular IT projects?

VI. Q&A / Break (1:45pm – 2:00pm)

VII. Support Vector Machines (2:00pm – 3:00pm)

  • The maximal margin classifier
  • The support vector classifier
  • Kernels and SVMs
  • Performance comparison to other classification methods

VIII. Q&A / Break (3:00pm – 3:15pm)

IX. Workshop: Let’s build and tune SVMs! (3:15pm – 4:45pm)

X. Workshop Feedback, Presentation from Winning Model and Day 2 Wrap Up (4:45pm – 5:00pm)

Day 3

+

I. Market Basket Analysis and Affinity Analysis I (9:00am – 9:45am)

  • What is association rule mining?
  • What is the business case for market basket analysis?
  • Support, lift and confidence
  • Visualizing market basket results

II. Q&A / Break (9:45am – 10:00am)

III. Workshop: Let’s use arules to perform MBA on supermarket data (10:00am – 11:30am)

IV. Introduction to Kaggle Competitions (11:30am – 12:15pm)

  • Kaggle overview
  • Kaggle competition strategies
  • Private and public LB
  • Team merging

V. LUNCH (12:15pm – 1:15pm)

VI. Workshop: Team Activity (1:15pm – 4:45pm)

During this capstone team activity, course participants will enrol in a live Kaggle competition. With the aim of achieving a top 50% leaderboard ranking by the end of the day, the full data science process will be implemented. Toward the end of the task, a strategy for continued learning and success in the competition will be discussed.

VII. Workshop Feedback and Course Wrap-up (4:45pm – 5:00pm)

$2,400
861

Sydney

Altis Consulting
Level 6, 219 Castlereagh Street
Sydney NSW 2000

24th to 26th July 2017

Single Rate $2,400*

* prices exclude GST

Meet our Instructor

Isaac Reyes

Isaac Reyes

Isaac is Head of Data Science at Altis Consulting. A passionate data science educator, Isaac has lectured in analytics and statistical theory at the Australian National University, AIM and the University of Canberra.

Isaac shared his vision for Data Science with perhaps the biggest teaching stage of them all – TEDx. “Speaking about the intersection of Data Science and world issues at a TED event was something that I’ve always wanted to do. My TED talk focused on how we can use Data Science to measure how much we really care about the issues that matter.”

In his previous roles, Isaac has worked as a Data Scientist at leading data consulting firms including Datapharm, Quantium and PricewaterhouseCoopers. He holds a Master of Statistics from the Australian National University and has over 3,000 hours of combined teaching experience. A top Kaggler, his vision is to train 1,000 Data Scientists by 2018.