Predictive Analytics - Speaker Deck

Slide 1

Slide 1 text

Slide 2

Slide 2 text

Slide 3

Slide 3 text

Slide 4

Slide 4 text

© 2023, Amazon Web Services, Inc. or its affiliates. Agenda • Data and digital transformation • Becoming data- and insight-driven organizations • Goals for data strategies • Predictive analytics • Prerequisites for predictive models • Model types • Common predictive algorithms 4

Slide 5

Slide 5 text

© 2023, Amazon Web Services, Inc. or its affiliates. Helen Beal 5 Herder of Humans @helenhappybee PURPOSE: Bringing Joy to Work Helen Beal is chair of the Value Stream Management Consortium, co-chair of the OASIS Value Stream Management Interoperability Technical Committee, and chief ambassador at DevOps Institute. She also provides strategic advisory services to DevOps and VSM industry leaders. Helen is the author of the annual State of VSM Reports from the VSMC and the State of Availability Report from Moogsoft. She is a co-author of the book about DevOps and governance, Investments Unlimited, published by IT Revolution. She is a DevOps editor for InfoQ, and hosts the Day-to-Day DevOps webinar series for BrightTalk and speaks on DevOps and value stream-related topics at a wide variety of industry conferences and at corporate events. She regularly appears in TechBeacon’s DevOps Top100 lists and was recognized as the Top DevOps Evangelist 2020 in the DevOps Dozen awards and was a ﬁnalist for Computing DevOps Excellence Awards’ DevOps Professional of the Year 2021. She serves on advisory and judging boards for many initiatives including Developer Week, DevOps World, JAX DevOps, and InterOp.

Slide 6

Slide 6 text

© 2023, Amazon Web Services, Inc. or its affiliates. About DevOps Institute 6 We drive human transformation in the digital age. DevOps Institute is a professional association and certiﬁcation authority that prepares people and organizations to succeed in building the processes and culture to support the future of IT. We are also a learning destination and community for technology practitioners and leaders looking to continuously learn about the IT technologies and processes that drive enterprise transformation.

Slide 7

Slide 7 text

© 2023, Amazon Web Services, Inc. or its affiliates. Data and digital transformation 7 DRIVER GOALS ENABLERS CHALLENGES OUTCOMES Digital transformation Exploiting internet-based opportunities to maximize organizational performance • DevOps • Cloud • Culture • Architecture Enhanced customer experience through speed of delivery Data transformation Delivering value through greater understanding, alignment, and actioning of digital and offline data • DataOps • Analytics • Omnichannel • Fragmentation Intelligence and insights unlocked to optimize decision making

Slide 8

Slide 8 text

Slide 9

Slide 9 text

Slide 10

Slide 10 text

Slide 11

Slide 11 text

© 2023, Amazon Web Services, Inc. or its affiliates. Four types of data analytics 11 “What happened?” Descriptive 1 “Why did this happen?” Diagnostic 2 “What should we do next?” Prescriptive 3 “What might happen in the future?” Predictive 4

Slide 12

Slide 12 text

© 2023, Amazon Web Services, Inc. or its affiliates. Four types of data analytics 12 “What happened?” “Why did this happen?” “What should we do next?” “What might happen in the future?” Descriptive 1 Diagnostic 2 Prescriptive 3 Predictive 4

Slide 13

Slide 13 text

© 2023, Amazon Web Services, Inc. or its affiliates. Predictive analytics 13 Calculating the likelihood of outcomes + = Computer modelling Data mining Machine learning Historical data Likely result of strategy change

Slide 14

Slide 14 text

© 2023, Amazon Web Services, Inc. or its affiliates. Use cases 14 Product manager CFO Cloud engineer How successful will this product roll out be? How might changing this feature within the product interface affect the customer’s likelihood to convert? What effect will this maintenance action have on the cloud environment? Is there a vulnerability in our system that could allow for an attack or breach? What does the future health of our business look like if we increase sales in this area and reduce expenses here? How will these investments impact shareholder value?

Slide 15

Slide 15 text

© 2023, Amazon Web Services, Inc. or its affiliates. 15 Content recommendation Customer lifetime value Customer segmentation Next best action Predictive maintenance Product propensity Quality assurance Risk modelling Sentiment analysis Upselling and cross selling Predicting buyer behavior Fraud detection Healthcare diagnosis Virtual assistance Campaign management Volume prediction Staffing and resourcing Stock and seasonality

Slide 16

Slide 16 text

© 2023, Amazon Web Services, Inc. or its affiliates. Predictive analytics examples 16 • Predict buying behavior • Recommender systems • Stock predictions • Warehouse: Customer locations • Personalized recommendations • Content development • Operations optimization • Customized marketing/trailers • Predict race outcomes • Build dynamic models of the race cars • Aerodynamics testing • Overtaking likelihood • Parts failures

Slide 17

Slide 17 text

Slide 18

Slide 18 text

© 2023, Amazon Web Services, Inc. or its affiliates. Predictive Analytics Maturity Model 18 Source: Econsultancy / RedEye Predictive Analytics Report £ Time Analysis limited to descriptive analytics Data sources disparate or limited No dedicated resource and siloed business culture Competency in descriptive and diagnostic analytics Data access limitations still exist Dedicated headcount Business culture focused on gaining beneﬁts from actionable insights Competency in predictive and prescriptive analytics Speciﬁc analytical environment available with single data repository New and additional data sources being sought Data-driven decision making prominent throughout the business Increasing competency, usage and optimization Starting out = no budget Developing = implementation is a business priority Strategic = functioning predictive analytics capabilities

Slide 19

Slide 19 text

© 2023, Amazon Web Services, Inc. or its affiliates. Prerequisites for predictive models 19 Appropriate sources of data Automation and machine learning Meeting business objectives Data cleanliness and usefulness Data can also be from third parties e.g. from data marketplaces

Slide 20

Slide 20 text

Slide 21

Slide 21 text

Slide 22

Slide 22 text

Slide 23

Slide 23 text

Slide 24

Slide 24 text

Slide 25

Slide 25 text

Slide 26

Slide 26 text

© 2023, Amazon Web Services, Inc. or its affiliates. GBM GLM Random forest K-means Prophet Common predictive algorithms 26 • Structural data in a table • Linear train more quickly • Nonlinear are better optimized for the problems they face (often nonlinear) MACHINE LEARNING • Structural data in a table • Linear train more quickly • Nonlinear are better optimized for the problems they face (often nonlinear) DEEP LEARNING

Slide 27

Slide 27 text

Slide 28

Slide 28 text

© 2023, Amazon Web Services, Inc. or its affiliates. Generalized linear model (GLM) for two values 28 Log of odds ratio Linear predictor (environmental suitability score) Link function Link function for binomial data = logit = log of odds ratio = log Probability of presence Probability of absence

Slide 29

Slide 29 text

Slide 30

Slide 30 text

Slide 31

Slide 31 text

Slide 32

Slide 32 text

© 2023, Amazon Web Services, Inc. or its affiliates. 32 Key Takeaways ● Digital transformation is driving data transformation ● Extracting value from data requires analytics to derive actionable insights ● There are different types of analytics Predictive analytics provide future driven insights for success ● Predictive analytics calculate the likelihood of outcomes ● Combine historical data with modelling, mining and machine learning ● Augment your data with third party date for more accuracy ● Use cases for personas across all industries ● Models include classiﬁcation, clustering, forecast, outliers, time series ● Algorithms in machine learning: random forest, GLM/GBM… Why What How