Student Dropout Prediction Pitch

https://nacollab-public.sharepoint.com/_layouts/15/guestaccess.aspx?guestaccesstoken=XL6B2rf%2bRn36G6F3w5s7ThDAUkoGlkTVMf4aeUsaQzY%3d&docid=001ac706c21624d489ab11a93c6f3614b

Machine LearningStudent Dropout Prediction
1.0x

Student Dropout Prediction Pitch

Created 2 years ago

Duration 0:07:48
lesson view count 72
https://nacollab-public.sharepoint.com/_layouts/15/guestaccess.aspx?guestaccesstoken=XL6B2rf%2bRn36G6F3w5s7ThDAUkoGlkTVMf4aeUsaQzY%3d&docid=001ac706c21624d489ab11a93c6f3614b
Select the file type you wish to download
Slide Content
  1. Azure Machine LearningStudent Dropout Prediction

    Slide 1 - Azure Machine LearningStudent Dropout Prediction

    • David Brown
    • Solution Sales Director
    • May 5, 2015
  2. Slide 2

    • About Us
    • Drive customer value with Predictive Analytics on the Microsoft platform.
    • Our objective is to make analytics accessible to institutions of all sizes across our verticals. Our team specializes in creation of analytical practices to help institutions grow and scale.
    • We are Seattle-based company with 25 data engineers and scientists that have helped dozens of customers improve their businesses. We were founded in 2011.
    • We are a Microsoft partner that develops solutions on Azure ML, Office 365, and Power BI.
    • Demand Forecasting, Decision Modelling, Resource forecasting, systems integration and creating more profitable customers
    • Solution Sales Director: David Brown, 425-283-6842, davidb@nealanalytics.com
    • http://www.nealanalytics.com
    • OUR PROCESS
    • Prove ROI
    • Broadly Engage Leaders
    • Production
    • Our Mission
    • Our Company
    • Edu., Retail, Energy
    • Partnerships
    • Our Focus
    • 2
  3. Challenges and Opportunities

    Slide 3 - Challenges and Opportunities

    • 3
    • Massive volumes of student data allow for more in-depth analysis
    • Universities are challenged to see through all this data in order to understand what’s really happening
    • Opportunities are missed or never even realized
    • Machine Learning and Predictive Analytics
    • Gain clear and detailed understanding of your student’s drivers
    • Predict outcomes of business decisions
    • Plan for effects of external influencers
    • Understand and resolve student issues before they happen
  4. Competitive Advantage

    Slide 4 - Competitive Advantage

    • Analytics
    • Azure Machine Learning
    • Machine Learning
    • Optimization
    • What’s the best that can happen?
    • Predictive Modeling
    • What will happen next?
    • Forecasting
    • What if these trends continue?
    • Statistical Analysis
    • Why is this happening?
    • Alerts
    • What action is needed?
    • Query & Drill-Down
    • Where exactly is the problem?
    • Ad Hoc Reports
    • How many? How often? Where?
    • Standard Reports
    • What happened?
    • Access & Reporting
    • Degree of Intelligence
    • Return on Investment
    • Current State (Excel)
    • Wants
    • Needs
  5. Slide 5

    • 5
    • Neal Analytics Modelling: Spectrum Approach
    • Black Box Models
    • Black Box
    • Stores
    • SKU’s
    • Routes
    • Clear Box Models
    • External Data Sources
    • Internal Data Sources
    • External Data Sources
    • Internal Data Sources
    • Forecast Demand Accurately
    • Sharpen Competitive Advantage
  6. Internal transactional data can be leveraged, and then joined with external data to provide deeper insights.

    Slide 6 - Internal transactional data can be leveraged, and then joined with external data to provide deeper insights.

    • Black box machine learning models to detect patterns
    • Data visualization and “clear box” models to tease apart what is going on
    • Combining black box and“clear box” models for datascience “stereoscopic vision”
    • Federation of “social listening”and other data sources withtransactional data, to extend organizational insights
    • Data
  7. Decision Profit Model

    Slide 7 - Decision Profit Model

    • External Data Sources
    • Internal Historical Data
    • Decision Profit Model Overview (Retail Example)
    • 7
    • Internal Variables
    • Internal controls are areas of your business where managers make decisions to control results like:
    • Promotion Plans
    • Advertising Plans
    • Distribution Plans
    • Product Plans
    • Domain Expertise
    • External Variables
    • External influencers are uncontrollable acts which influence your business like:
    • Weather (if applicable)
    • Key sales holidays
    • Socio-economic Factors
    • Product Seasonality
  8. Decision Profit Model

    Slide 8 - Decision Profit Model

    • External Data Sources
    • Internal Historical Data
    • Decision Profit Model Overview (Edu. Example)
    • 8
    • Internal Variables
    • Internal controls are areas of your business where managers make decisions to control results like:
    • School Sporting Events
    • Academic Reputation
    • Dorm Accommodations
    • Course Availability
    • Degree Offerings
    • External Variables
    • External influencers are uncontrollable acts which influence your business like:
    • Socio-economic Factors
    • Enrolment Seasonality
    • Public Funding
    • Family Life
  9. Student Attrition Case Study: Australia

    Slide 9 - Student Attrition Case Study: Australia

    • 9
    • Top 5 Performing Universities (based on Annual % of Students Lost )
    • The University of Melbourne (6%)
    • The Australian National University (7%)
    • Monash University (7%)
    • University of New South Wales (8%)
    • University of Sydney (8%)
    • Cost of Student Attrition for Top 5 Performing Universities
    • Student Enrolments x Attrition % x Life Time Value of Student
    • Conservative estimate that the Life Time Value of a Student is $50K (both Domestic & International though we know International are significantly higher). Around $40K for teaching over 3-4 years plus a sum generated through additional University revenues over their time.
    • University of Melbourne - $142 million (47,561 Students x 6% Attrition Rate x $50K LTV of Student)
    • The Australian National University - $67 million (19,313 Students x 7% Attrition Rate x $50K LTV of Student)
    • Monash University - $221 million (63,338 Students x 7% Attrition Rate x $50K LTV of Student)
    • University of New South Wales - $202 million (50,613 Students x 8% Attrition Rate x $50K LTV of Student)
    • University of Sydney - $204 million (51,168 Students x 8% Attrition Rate x $50K LTV of Student)
  10. Excel BI   Power BI	Azure ML

    Slide 10 - Excel BI Power BI Azure ML

    • DEMO
  11. Offering

    Slide 17 - Offering

    • 17
  12. Student Attrition Case Study: Australia

    Slide 18 - Student Attrition Case Study: Australia

    • 18
    • 5 Year Change in Attrition for Domestic & International Students
    • The University of Melbourne (-0.4%) $128 million. 0.4% reduction equates to $14 million.
    • The Australian National University (-1.6%) $52 million 1.6% reduction equates to $15 million.
    • Monash University (0.9%) $225 million 0.9% increase equates to $4 million.
    • University of New South Wales (0.7%) $220 million 0.7% increase equates to $18 million.
    • University of Sydney (0.5%) $217 million 0.5% increase equates to $13 million.
    • How much would this impact your institution?
  13. 19

    Slide 19 - 19

    • Proof of Profit (PoP) Methodology
    • Competitive Model
    • Existing forecasting model
    • Neal Analytics model
    • The actual result
    • Methodology
    • Identify business scenario to be explored
    • Identify external variables as explanatory variables
    • Train models with historical data (3 years min.)
    • Evaluate 2014 data with legacy system
    • Document 2014 outcome
    • Determine lift from new methodology
    • “Seeing is Believing”
    • Sales Forecast Error
    • PoP performance is extrapolated to indicate solution performance.
    • Once the improvement in forecasting is estimated, an ROI business case for the final solution is created.
    • Legacy System
    • New System
    • Lift
  14. 20

    Slide 20 - 20

    • Contact Me
    • David Brown
    • davidb@nealanalytics.com
    • 1 425 283 6842
  15. Approach

    Slide 21 - Approach

    • 21
    • Business Critical Success Factors (CSF)
    • Key Performance Indicators (KPI)
    • Catalog Internal/External Variables
    • Inventory database systems
    • Evaluate available data
    • Define solution
    • Identify subset of requirements
    • Determine success thresholds
    • Work with historical data only
    • Compact Excel-based deliverable
    • Fast (4 to 6 weeks) and inexpensive
    • Segment project into phases
    • Deliver high ROI phases first
    • Reduce risk with pilots
    • Continuous improvement
    • Track received value
    • Gather Requirements
    • Proof of Concept
    • Production System
  16. 22

    Slide 22 - 22

    • Typical Proof of Profit Process
    • rivers
    • Industry overview
    • Sources of competitive advantage
    • Competitive environment
    • Critical Success Factors (CSF)
    • Key Performance Indicators (KPI)
    • Strategy
    • Critical business processes
    • Definition of “Success”
    • Evaluate data environment
    • Applications
    • Databases
    • Data structures
    • Data quality and gaps
    • External variables
    • User requirements
    • Usability
    • Accessibility
    • Support
    • Receive data
    • Production databases
    • Flat files
    • External data sources
    • Data cleansing
    • Remove duplicates
    • Fix formatting
    • Fix consistency errors
    • Address missing data
    • Create data structure
    • Combine data
    • Create relationships
    • Document database
    • Data upload
    • Choose Azure storage model
    • Transfer the data to Azure
    • Exploratory Data Analysis
    • Descriptive statistics
    • Evaluate seasonality
    • Data visualization
    • Modeling
    • Develop strategy
    • Apply transforms
    • Train algorithms (5 years of data)
    • Test algorithms
    • Tune algorithms
    • Algorithm selection
    • Choose POC algorithms
    • Document process
    • Develop interface
    • Build in Microsoft Excel
    • Visualize with Power View
    • Get user feedback
    • Revise interface
    • Document interface
    • Evaluate models
    • Access test year data
    • Benchmark legacy system
    • Benchmark POC system
    • Calculate lift for POC system
    • Build business case
    • Document business drivers
    • Generate value matrix
    • Create 5-year cash flows
    • Calculate NPV, IRR and ROI
    • 5 days*
    • 15 days*
    • 5 days*
    • 10 days*
    • * Represents calendar days opposed to 8 hour work days
    • Solution Definition
    • Data Engineering
    • Modeling
    • Solution/Evaluation
  17. Is passionate about solving tough business problems with data. His great curiosity has led to new insights and predictive models in energy systems, social networks, supply chain networks, and marketing. An innovator at heart, Zach has worked at the intersection of business and technology for over eight years, leading analytics projects at companies such as Dong Energy, Vestas, and Siemens. Originally an Oregonian, Zach has lived and worked in all four corners of the United States as well as France and Denmark. While living in Denmark, Zach harnessed his background in engineering and business to help develop some of the largest offshore wind energy projects in the world. When he is not focusing on tackling new data challenges, Zach is typically exploring the great outdoors of the Northwest by foot, kayak, or paddle board.

    Slide 23 - Is passionate about solving tough business problems with data. His great curiosity has led to new insights and predictive models in energy systems, social networks, supply chain networks, and marketing. An innovator at heart, Zach has worked at the intersection of business and technology for over eight years, leading analytics projects at companies such as Dong Energy, Vestas, and Siemens. Originally an Oregonian, Zach has lived and worked in all four corners of the United States as well as France and Denmark. While living in Denmark, Zach harnessed his background in engineering and business to help develop some of the largest offshore wind energy projects in the world. When he is not focusing on tackling new data challenges, Zach is typically exploring the great outdoors of the Northwest by foot, kayak, or paddle board.

    • BS Bioengineering – University of California Berkeley
    • MBA – Georgia Institute of Technology
    • MS Mechanical Engineering – Georgia Institute of Technology
    • Sample Personnel (1/2)
    • 23
    • Zach PerkelLead Data Scientist
    • B.A. – Political Science University of California Davis
    • MBA – Southern Methodist University
    • MS – Predictive Analytics Northwestern University
    • Bill has 20 years of Information Technology experience earned in the public and private sector. In the public sector, he worked for the U.S. Department of State at embassies in Saudi Arabia, Lebanon, Germany, Nigeria and Colombia. While in this role, he developed predictive analytics solutions for fraud detection. He also worked closely with the oil industries in these countries and served as the Chief Marketing Officer for the Lagos Oil Club in Nigeria. In the private sector, he worked for KPMG, Microsoft and Wipro Technologies. While with Microsoft Consulting Services, Bill provided services to dozens of energy companies delivering infrastructure and security solutions. He attended Rice University’s Energy Management program and has a certificate in photovoltaics from Stanford University. While working for Wipro Technologies, he led a team that developed a human capital management solution for offshore oil and gas. This project was done in collaboration with Rice University and several large oil companies. Bill has been published in the Journal of Petroleum Technology. He is passionate about demonstrating the business value of predictive analytics and machine learning technologies to his customers in the energy industry.
    • Bill BarnaPractice Director
  18. Sample Personnel (2/2)

    Slide 24 - Sample Personnel (2/2)

    • 24
    • Is a modeller with 25 years experience working across disciplines (molecular biology, electrical engineering, marketing mix, price elasticity, & promotion baselining) across computing platforms (mainframe, micro, cloud, & embedded devices) and across analytical platforms (SAS, SQL Server, R, SYTAT/SYGRAPH, & MicroTSP). Bill develops pragmatic models which opportunistically exploit available information. A believer that opportunity is always manifested in numbers, Bill is a hands on problem solver eager to meet users, understand challenges, and to analytically pin down breakthrough insights, and data visualizations. Bill has worked on (as well as managed) a broad spectrum of technical teams (scrum and waterfall software, embedded systems, data science, litigators, & house counsel, among others) and built a world-wide IT system for Hewlett-Packard’s legal department without headcount or budget. This system led to an overall increase invention disclosure of 5,000% over the system’s life.
    • Bachelor’s – Finance, Michigan State University
    • Master’s – Business Administration, Michigan State University
    • Ph.D. – Quantitative Marketing, Michigan State University
    • Ph.D. – Minors in Electrical Engineering, Evolutionary Ecology, Econometrics & Statistics, & Minor in Accounting Information Systems, Michigan State University
    • Bachelor’s – Management Information Systems University of South Florida
    • Microsoft Certified BI Sales Specialist – SQL 2012 Microsoft Corporation
    • Bill MeadeData Science Director
    • Drawing on nearly 15 years of business development experience with companies like Hyperion Solutions (acquired by Oracle), IBM’s Information Management Division, and Extended Results (acquired by TIBCO Spotfire), Greg is uniquely positioned to perform as a trusted advisor to our customers. His passion for business intelligence is only equaled by his commitment to quality and customer satisfaction. Greg attended the University of South Florida (Management Information Systems), began his professional career as a software developer, has travelled to 27 different countries, enjoys spending his weekends cheering for his 3 children as they compete in a variety of sports and is fluent in Spanish.
    • Greg Gomez
    • VP of Sales