Machine Learning - why the hype and how it does its magic
- 1. Machine Learning Why the hype & how it does its magic DevNexus Atlanta By: Amir Charania on Feb 23rd 2017
- 2. Confidential & Proprietary to Daugherty Business Solutions. 2 Agenda Machine Learning – Why the Hype? Machine Learning - Fundamentals Machine Learning - Framework and Process Machine Learning in Action! Model Building, testing, deploying Machine Learning Use Cases
- 3. Confidential & Proprietary to Daugherty Business Solutions. 3 Section 1 Machine Learning – Why the Hype?
- 4. Confidential & Proprietary to Daugherty Business Solutions. 4 Source: http://www.nytimes.com/2004/03/01/business/microsoft-amid-dwindling-interest-talks-up-computing-as-a-career.html?_r=0
- 5. Confidential & Proprietary to Daugherty Business Solutions. 5 Thought leaders like Gartner identify it as a trend Gartner defines a strategic technology trend as one with the potential for significant impact on the organization. Factors that denote significant impact include a high potential for disruption to the business, end users or IT, the need for a major investment, or the risk of being late to adopt Gartner Top 10 Strategic Technology Trends 2016
- 6. Confidential & Proprietary to Daugherty Business Solutions. 6 Gartner 2015 – Hype Cycle Machine Learning It’s made it to the Gartner Hype Cycle
- 7. Confidential & Proprietary to Daugherty Business Solutions. 7 And the hype increased in 2016! Machine Learning Gartner 2016 – Hype Cycle
- 8. Confidential & Proprietary to Daugherty Business Solutions. 8 Major players have launched ML platforms Less than a year ago!
- 9. Confidential & Proprietary to Daugherty Business Solutions. 9 • Increasing data volumes • Low storage costBig Data • Zero startup (hardware & software) cost • Reduced startup time • Pay for what you use • Deployment is a snap Cloud • Microsoft, Amazon & Google now have platforms for Machine Learning • Ability to play with End-to-End tools for free! Major Players Entering Market • Ability to learn from the best in the world • Availability of open source data sets and tools to experiment and learn MOOC Multiple factors are creating the perfect storm for ML Democratization of Predictive Analytics • Market is ripe for mainstream adoption • Companies with access to data, talent & the right strategy are poised to win
- 10. Confidential & Proprietary to Daugherty Business Solutions. 10 But reality is that analytics is easier said than done Source: MITSloan Management Review: Findings from 2016 Data & Analytics Global Executive Study and Research Project Managing with analytics is now a mainstream idea, though not a mainstream practice
- 11. Confidential & Proprietary to Daugherty Business Solutions. 11 That trend is only increasing …few companies have a strategic plan for analytics or are executing a strategy for what they hope to achieve with analytics Source: MITSloan Management Review: Findings from 2016 Data & Analytics Global Executive Study and Research Project
- 12. Confidential & Proprietary to Daugherty Business Solutions. 12 Section 2 Machine Learning - Fundamentals
- 13. Confidential & Proprietary to Daugherty Business Solutions. 13 ML falls within the Predictive Analytics step of the Analytics Escalator DESCRIPTIVE • Reports • Dashboard • Business Intelligence DIAGNOSTIC • Queries • Statistical Analysis • Cubes/OLAP Tools • Machine Learning • Predictive models • Forecasting • Optimization • Planning PREDICTIVEPRESCRIPTIVE
- 14. Confidential & Proprietary to Daugherty Business Solutions. 14 What is Machine Learning “The goal of machine learning is to build computer systems that can adapt and learn from their experience” - Tom Dietterich Thomas G. Dietterich is Emeritus Professor of computer science at Oregon State University. He is one of the founders of the field of machine learning.
- 15. Confidential & Proprietary to Daugherty Business Solutions. 15 Example 1 Based on the size of the Tumor, predict whether the Tumor is Malignant or Benign Tumor Size Result 3.4 Malignant 4.2 Benign 1.2 Benign 2.3 Benign 5.2 Malignant 4.7 Malignant Tumor Size Result 1.9 ? Classification Problem Feature Label Observations
- 16. Confidential & Proprietary to Daugherty Business Solutions. 16 Example 2 Based on the size of the house, predict how much the house will sell for? House Size Sales Price 2400 sq. ft. $230,000 3200 sq. ft. $410,000 1800 sq. ft. $167,000 2100 sq. ft. $225,000 3000 sq. ft. $350,000 2800 sq. ft. $310,000 House Size Sales Price 2200 sq. ft. ? Regression Problem
- 17. Confidential & Proprietary to Daugherty Business Solutions. 17 Section 3 Machine Learning – Framework & Process
- 18. Confidential & Proprietary to Daugherty Business Solutions. 18 Process for solving real world problems Ask the right question Frame the question so that Machine Learning can be applied Apply Machine Learning Reframe the ML answer to real-world Extract Data Develop Model Deploy Model Evaluate Model Performance Define Target Metric Extract Derived Features Select FeaturesFit Model Evaluate Models
- 19. Confidential & Proprietary to Daugherty Business Solutions. 19 Machine Learning Framework Input Data Function Output Machine Learning Algorithm y = f(x) output prediction function Feature - House size
- 20. Confidential & Proprietary to Daugherty Business Solutions. 20 Machine Learning Process Training Tumor Size Result 3.4 Malignant 4.2 Benign 1.2 Benign 2.3 Benign 5.2 Malignant 4.7 Malignant Training Data Set Training Features Training Training Label Learned Model Testing Tumor Size Result 1.9 ? Testing Features Learned Model Prediction
- 21. Confidential & Proprietary to Daugherty Business Solutions. 21 What is NOT Machine Learning Find the average sales price of the home Calculating the # of Malignant vs Benign tumors ML Goal: Build predictive models
- 22. Confidential & Proprietary to Daugherty Business Solutions. 22 Section 4 Demo Time Machine Learning in Action! Model Building, testing, deploying
- 23. Confidential & Proprietary to Daugherty Business Solutions. 23 Use Cases of Machine Learning Retail • Customer Churn • Predicting Customer LTV • Cross-selling & Recommendations Algorithms • Market Basket Analysis Hospitality • Inventory Management/Dynamic Pricing Airline • Proactive Equipment Maintenance Financial Services • Fraud Detection • Credit Risk Source: https://www.kaggle.com/wiki/DataScienceUseCases
- 24. Confidential & Proprietary to Daugherty Business Solutions. 24 Source: Microsoft - Dig Deep with Azure Machine Learning Here is how you select the Machine Learning Algorithm
- 25. Confidential & Proprietary to Daugherty Business Solutions. 25
- 26. Confidential & Proprietary to Daugherty Business Solutions. 26 Thank you! Thank you!