Note 1 Flashcards
What is machine Learning
Machine Learning is about Predicting Results based on Incoming Data
Steps in Full Machine Learning Project
- Data Collection
- Data Modeling
2.1 Problem definition
2.2 Data
2.3 Evaluation
2.4 Features
2.5 Modelling
2.6 Experiments - Deployment
Types of data in machine learning
Supervised
Unsupervised
Reinforcement Learning
Supervised Data
Data we receive already have categories (Rows and columns) and text data
Example - Function is right or wrong
Classification - 2 categories
Regression - Multiple Categories
Unsupervised Data
Data which Doesn’t have labels
Clusters and association Rule Learning
Clusters is nothing but grouping Operations
Association Rule is like what customer will buy in the Future
Reinforcement Learning
Real time Learning
Teaching Machine through trial and error
(Through reward and punishment)
Different categories in Reinforcement Learning
- Neural Networks
- Decision Trees
- SVM
- KNN
Learn from the data they receive and Predict
Classification
is there apple or orange
it draw lines to decide this is apple and this is orange
Regression
Predict stock Price
Hiring engineer based on Inputs (Year of exp, Age, what type of computer they have) - Labels
Association Rule Learning
Where we associate Different things to predict what customer paraphs might buy in the feature.
How Machine Learning algorithm works
Input (All Ingredients)
Output (Ideal - Assuming Output)
It look at the input and look at the output and it try to figure out the set of instructions in-between the two
you may have 100 or 1000 of attempts
It may take 1000’s of times to find right instructions.
Data Analysis
Looking at set of data and understanding it by comparing different examples, Different Features, Visualizations like graphs comparing them
Data Science
is experimenting with different data or set of data finding actionable insights within it.
or
Building a Machine learning model.
Steps in Machine Learning Framework
1)What Problem we are trying to solve?
(It is supervised or unsupervised problem)
2)Data
(Structured Data - Row and columns (Excel)
(Unstructured Data - Image / Audios
3)Evaluation
(What Defines success for us)
(Something to Aim)
Example :
House data — Machine Learning model — House price = Predicted Price 4,97,000 and Actual Price 5,00,000 } How accurate 90 - 95 % Accurate
4)Features
(what do we already know about the data)
Example :
Heart Disease — (Features) Body weight, Blood Pressure, Chest pain
5)Modeling :
(Based on our problem and data , what model should we use?)
Example :
Problem 1 — Model 1
Problem 2 — Model 2
Some model works better
Right model for the right kind of problem
6)Experimentation :
(How could we improve / what can we try next?)
Example :
Attempt 1 — Fail
Attempt 2 — Fail
Attempt 3 — Partial Success
Attempt 4 — complete Success
Detailed Explanation on Problem Identification (Step 1)
Page No : 13
Problem Identification (Definition)
Main types of Machine Learning
- Supervised Learning
- Reinforcement Learning
- unsupervised Learning
- Transfer Learning
Supervisad Learning
- You have Data and Labels
- Machine Lerning: Algo tries to use the Date to predict the label If it guesses, wrong the algo Correct itself and tries again. This active Correction, is called Supervised
Data
(Collectim of all Records)
Label
(output Results)
|
|
output
(instructions)
Supervised Learning Repeat the process over and over time to get better
- Classification
- Regression
Classification:
- Binary Classification - two option
- Multiclass Classification - More than two option
Regression
- Trying to predict the No.
- Refer to us Continuous No.
- No. may go up or Down
Example:
- No. of Room
- how many people will buy
Example:
How Many people will by this app
How much will this house sell for?
Unsupervised Learning:
-Purchase history
ID Purchase 1 Purchase 2
1 Sunglasses Singlet (Summer)
2 Jacket Show boots (Winter)
3 Sun Screen Beach towel (Summer)
Sent Promotion for Next Summer
who is intresting in Purchasing for Summer
Do Classificationnalysis
- I don’t have output but I have Input
Transfer Leaning:
- I what dog Bread appear in the photo (Predict) Cor Model Machine learning algo use its foundation, Platform and apply it to the day photo
- I think my Problem may similar to Something else. Leverage what existing Machine Learning Model has learned.
Reinforcement Leaning:
- I having a Computer program, platform Some action within Define Space and Rewarding it whom it doing well and Punishing it when doing wrong
+ Example: Teaching Machine Learning Algo playing chess
win +1
Lose -1
What is Supervised Learning
Page No : 13
Supervisad Learning
- You have Data and Labels
- Machine Lerning: Algo tries to use the Date to predict the label If it guesses, wrong the algo Correct itself and tries again. This active Correction, is called Supervised
Data
(Collectim of all Records)
Label
(output Results)
|
|
output
(instructions)
Supervised Learning Repeat the process over and over time to get better
- Classification
- Regression
Classification:
- Binary Classification - two option
- Multiclass Classification - More than two option
Regression
- Trying to predict the No.
- Refer to us Continuous No.
- No. may go up or Down
Example:
- No. of Room
- how many people will buy
Example:
How Many people will by this app
How much will this house sell for?
What is Unsupervised Learning
Page No : 14
Unsupervised Learning:
-Purchase history
ID Purchase 1 Purchase 2
1 Sunglasses Singlet (Summer)
2 Jacket Show boots (Winter)
3 Sun Screen Beach towel (Summer)
Sent Promotion for Next Summer
who is intresting in Purchasing for Summer
Do Classificationnalysis
- I don’t have output but I have Input
What is Transfer Learning
Page No : 14
Transfer Leaning:
- I what dog Bread appear in the photo (Predict) Cor Model Machine learning algo use its foundation, Platform and apply it to the day photo
- I think my Problem may similar to Something else. Leverage what existing Machine Learning Model has learned.
What is Reinforcement Learning
Page No : 15
Reinforcement Leaning:
- I having a Computer program, platform Some action within Define Space and Rewarding it whom it doing well and Punishing it when doing wrong
+ Example: Teaching Machine Learning Algo playing chess
win +1
Lose -1
Detailed Explanation on Data (Step 2)
Page No : 15
Data
what data do we hove
Different types of, Data
- structured
- unstructured
- Structured
it will look Like Excel file (Rows & Columns), Medical Records, Customer purchase Transactions is called Structured Data - unstructured Data
(like images, Natural language text Phone calls, Videos, audio files) we turn them into No and Make them structured. - Static Data
This Data won’t Choose over time (.csv) (or) Table
The More Dato better Hi Resolt is - Streaming Data
Data which is Constantly changing overtime.
Example: Stock price change based on news headlines
Note: Most of the work in practice will start on Static Data and Machine learning (Data Analysis) Effects Show some insights you Move towords Streaming Data
Workflow
Static Data (csv) — Id (Jupyter) — Exploring Data (Data Analysis) (Pandas) — Visualization (MatPlotlib) — Machine Learning Model on the Data (Scikit learn) — Result
Detailed Explanation on Evaluation (Step 3)
Page No : 16
Detailed Explanation on Features in Data (Step 4)
Page No : 18
Detailed Explanation on Modeling (Part 1) - Splitting Data (Step 5)
Page No : 19
Detailed Explanation on Modeling (Part 2) - Choosing a Model (Step 5)
Page No : 21
Detailed Explanation on Modeling (Part 3) - Tuning (Step 5)
Page No : 23
Detailed Explanation on Modeling (Part 4) - Comparison (Step 5)
Page No : 24
Overfitting and Underfitting and Balanced
Page No : 25
Several reason why underfitting and overfitting can happen
Page No : 26
How to comparing models
Page No : 28
Detailed Explanation on Experimentation (Step 6)
Page No : 29
Tools we use
Anaconda — Pandas, Matplotlib, Numpy, tensorflow, pytorch, XGboost, Scikit learn, Catboost
Setup Needed (Like Code Editor, IDE’s, Notebook)
Page No : 31
How to capture information on python code
Input (“What is your name”)
Python Program to Print
Print (“Gowtham”)
Python : How to store input
Name = input (“What is your name?”)
Print (Name)
or
Print (“hello “ + Name)
Note : + is an combining operation
4 key things you should know while learning a language
Yo Term
Data Types
Action
Best Practices
what is Data Types?
Page No : 34
Power Symbol
Modular Symbol
Denomination
**
%
//
Operational Symbol
*
/
Action (Functions) terms in python
Print - Printing result
type - Showing Class (float, int,….)
Math operation
ABS function (Action)
Round Function
Page No : 35 and 36
Operation Procedure of Math
Page No : 36
Yo terms
Variables
Page No : 37
keywords
list
if
for
while
Dunder
__
Variable should not be created with Dunder
Expression Vs Statement
Page No : 37
Augmented Assignment Operator
Page No : 38
Strings
Long String
Page No : 38
String Concatenation
type Conversion
Escape sequence
Formatted String
Page No : 39 & 40
Formatted string
Page no : 40
String index or slicing
Page no : 41
String index (step over)
Page no : 42
Reverse slicing
Page no : 41
What is immutability in slicing
Page no : 41
Built in function
Page no : 41
What is len and starting no. in len
Page no : 42
What is built in function and what is method
Words with () are build in function
Words starting with “.” are method
How to use uppercase in python
.upper()
Page no : 42
Things we do with string
String can’t be changed but can be deleted or overwrite or create new
What is boolian
True or false
Page no : 42
What is type conversion
Page no : 43
How to comment on python
single line comment
Page no : 43
Basic password checker exercise
Page no : 43
How to get input in python
Page no : 43
