PowerPoint Presentation
Scope of the Course
Copyright By PowCoder代写 加微信 powcoder
How is ISE view of Analytics different from CS Data Science, and ?
CS View of Data Science: Face Recognition, Speech Recognition
: Use Software (e.g. SAS) to make Business Decisions
ISE View: Models for Transforming Data to Decisions
What kind of Background?
Computing: We expect students to have had some programing experience, although we will introduce students to Python and R
Statistics: Regression, hypothesis testing. We will go over other stat/machine learning concepts such as overfitting, generalizability and others
Optimization: Linear Programming, but we will introduce some Stochastic Programming models
Interfacing Data and Decisions?
Cross-sectional Data
Time Series Data
Spatio-Temporal Data
Modeling Paradigm
LEARNING ENABLED OPTIMIZATION
INTEGRATIVE ANALYTICS
Big Data and Big Decisions
Modern OR APPLICATIONS
with Big Data and Big Decisions
Integrative Analytics
Coordination of Sales, Marketing and Production (Cross-Sectional Data)
Inventory Control (Time Series Data)
Analytics of
Things (AoT)
Renewable Energy Integration
In this talk, we wish to bridge the gap between Big Data and Big Decisions. These are decisions of major, global consequences. To the extent that INFORMS is the home for end-to-end analytics, it has an edge in leading the Integrative Analytics agenda.
In connection with this leadership, I believe that a methodology based on “Learning Enabled Optimization” will do for Analytics what “Optimization Enabled Learning” has done for Statistical Learning
THE SPECTRUM of ANALYTICS
Statistical Learning
Integrative Analytics
of Optimization
Example Application: Advertising+Production Planning
Sales Projection based on Regression
Use Sales Projections (data) in Production Planning
How should one create a production plan when sales is forecast using regression ? (Cross-sectional data)
Example Application: Inventory Models
Consider the “R” data set called elecequip
10 years of Time Series Data
How should one create an optimal ordering policy when the data is a time series
Alternative types of mean-reverting processes
Example Application: Renewable Energy
Wind data exhibits spatio-temporal (correlated) data
Validated off-line simulation can be used for simulation as a proxy for data
How should one use simulation wind output in a unit-commitment cum dispatch model
Other types of Data:
What is “Diet Problem into the ?”
Modern data sets: all kinds of Data
Healthcare Costs Energy Usage
Nutrition Content College Debt
Transportation Costs etc..
The Analytics Problems of Today Ask that We Use Optimization Methodology which will put a Strong Foundation for Analytics
Ohio Banking Problem
State of Ohio, and Counties
New law (in 1979) A Bans can put branches in any county where the bank has a principal place of business and in any county adjacent to one in which it has a principal place of business.
Counties in Ohio (marked by the shapes in the map)
(1979) … A Bank can put Branches in any county where the bank has a principal place of business and in any county adjacent to one in which it has a principal place of business.
What is the minimum number of principal places of business, and in which counties should they be located to enable branches in all eighty-eight counties of Ohio?
HW 1 (For Teams of Two)
Using population data of Ohio counties from the data closest to the year prior to 2011, instantiate the Branch Location formulation discussed in the Tutorial by .
Using population data for 2020, instantiate the Branch Location formulation.
Compare the solutions, and discuss whether the ”glide path” property presented in the paper applies to the decisions which you obtained.
Suppose that in 2011 you had a crystal ball into 2020, and wanted to use that population data during the 2011 run. Formulate a model that enforces the glide path property for 2020.
程序代写 CS代考 加微信: powcoder QQ: 1823890830 Email: powcoder@163.com