Summary Georgia Institute Of Technology ISYE 6501; ISYE6510 Modules 1-10, Complete Latest Fall 2025-26.

Module 1

Ayics important types of questions r



Descriptive-what happened?
-




-




Predictive what
-

will happen? Uncertantity wh outcomes
-




Prescriptive -


What action(s) would be best?
-


More general questions


Modding
-


Means taking a real life situation and expressing the key parts of that situation in terms of math.

O Describe a real life situation mathematically
② Analyze the math
③ Turn mathematical answer back into real-life situation




se
strut
at



Cuttingstep
Reg Output Quality
-



ression
-




-



Optimization -




Missing Data
-

Etc .
-

Etc




What is Modeling?
-

Real life situation
pressed as math
model




This study source was downloaded by 100000900934077 from CourseHero.com on 10-08-2025 13:25:53 GMT -05:00





https://www.coursehero.com/file/251983907/Mod-1-Intropdf/

,Module
2Classification
: is




Putting things into categories
-




-



As simple as yes or no

Able to Differentiate
-

More than just 2 categories
-


Classification Questions
ex-Bankloan -
Income age assets credit
, , ,




Blue dots are those who paid back a loan
Red are those who did not (defaulted

draw this line to determine if want to loan
-
we
give a




But you can draw
-

an infinite number of lines between these so what do you do ?
,



classifier
higher
L Here is a version with 2 lines , called separators

lowerossifiers or classifiers.
Blue above and red below.
-




-


what about this applicant in yellow?
The lower classifer says approve higher
deny
-




,




What do we do?

, In want to chose from
general line that's further making mistakes
-



, a




-


Suppose we made this line to separate ,
its
pretty close to
a red point and blue point which
,
means that it almost
-




makes couple of mistakes in classification.
a

If the line or the points were a little different ,
the line

might misclassify them .




Now suppose this line , much further from
any points not
-

we use ,




so close to making mistakes Would take much
.

bigger error

in the data to cause misclassification .
That is more what for .
looking
-



we are




This data now has no way to perfectly separate between the blue and red points.
We need soft
classifier. that gives
One us
good
-


a

as separation possible
as ,
hard classifier
ratherthan a

that separates perfectly
-

This line minimizes the number of incorrectly classified
points. There are three (in yellow) incorrect points.
But, it close to of mistakes. (circled)
making a lot
-




is

, If we filt and slide that line like this we get a ,



few more mistakes fewertotal mistakes and near
,


Mistakes .




Trading off mistakes and near mistakes depending , on now


important we think each one is.




How important
something is?
cost of each mistake is not always equal.
-




In loan
example-cost of giving a loan that is not repaid is more than
denying a loan mistakenly.
-

.




-


How change def of the best separator?
-


Ex The more costly one type of bad decision is
.

,
the more we want to more the line away from it.


In this ex ,
we determine the cost of a defaulted
loan Cred dots) double the cost of
is
denying a


good loan ,
so we ift the line closer to the

blue points than the red points .


New applicant in yellow-status ambiguous,
-


is

it's between the blue+ red points , our new

cost-concious classifier will
suggest deny.