project cycle - summary ARTIFICIAL INTELLIGENCE.pdf

AI PC

1. AI -417

2. Ch-2
AI PROJECT CYCLE

3. INTRODUCTION TO AI PROJECT CYCLE
–
Project
AProjectreferstoasetofoperations
carriedoutwithgivenresourceswithina
specificscheduletoachieveadefined
objective.
ProjectCycle
ProjectCyclereferstothelifecycleofany
projectthatdescribesdifferentproject
stages,witheachstagebeingseparatefrom
oneanother&deliveringormeetingcertain
objectives.
Thedifferentstagesfrominitiationtothe
closureofaproject,togetherareknownasa
ProjectCycle.
#easy2learn with Jyoti

4. Five STAGES OF AN AI PROJECT CYCLE
#easy2learn with Jyoti

5. 1. Problem Scoping
In the Problem Scoping Stage, broadly the aim & Scope of the
Project undertaken are decided.
During this stage following things are decided-
✓Business objectives
✓Expected Outcome of the Project
✓Stakeholders’ expectation
✓Key resources & steps
✓Success Matrices
#easy2learn with Jyoti

6. 2. Data Acquisition
✓In this stage correct data in the right form must be fed to an AI
Project.
➢The data are acquired keeping in mind the Scope & parameters
of the Previous stage.
➢For example for an AI-based Insurance Project, the data will be
heavily numbers and figures, while for a social-media-based AI
Project or for providing security to a bank, the data may be
mostly in visual form.
➢Accurate
➢Reliable
#easy2learn with Jyoti

7. iii. Data Exploration
The Purpose of the third stage is to explore data so as to –
✓Choose the type of model for the project that can solve the
problem.
✓Clean & Normalize the data to standardize & correlate the data.
[ Clean & Normalize dirty Data requires data scientists to make
decisions on data they may not understand. like what to do with
missing data, incomplete data, and deviating data (called outliers ).]
Thus, during this phase after deciding about the type of models that
may serve –
+Data from multiple sources is aggregated into a format suitable for
the AI’s Project’s model.
+The form & structure of data is such chosen so that it is likely to
produce the desired outcome.
#easy2learn with Jyoti

8. iv. Modelling
In this stage
✓The selected Models (from the previous stage) are tested and
analyzed & the most suitable AI Model is chosen that matches the
requirement of the Project.
✓Once the most efficient model is chosen, an AI algorithm is
developed around it.
During this Stage, the Training data and Testing data are also decided.
1. Training Data –it is a dataset i.e. used to train an AI algorithm.
2. Testing Data – it is a set of observations used to validate the
developed models after training is complete.
#easy2learn with Jyoti

9. v. Evaluation
In the Evaluation stage, the developed
model is actually evaluated for accuracy
& performance using new data so as to
determine if the developed model is
deployable or not.
#easy2learn with Jyoti

10. + An AI project undergoes some phases from its initiation to closure, which together are
known as the Al Project Cycle.
+In the problem scoping stage, broadly the aim and scope of the project undertaken
are decided.
+In the data acquisition stage, the data are acquired considering the scope and
parameters of the previous stage.
+In the data exploration stage, the data are explored to choose the possible models and
clean & normalize data.
+In the modeling stage, the selected models are tested and analyzed, the most
suitable AI model is chosen, and Al-algorithms are developed around it.
+In the evaluation stage, the developed model is evaluated for accuracy and
performance using new data to determine if the developed model is deployable or
not.
#easy2learn with Jyoti

11. #easy2learn with Jyoti

12. PROBLEM
SCOPING
4W CANVAS

13. ✓Problem Scoping - we decide the aim & scope of the AI Project.
✓In this stage aim is to develop an AI-based solution for a problem –
firstly we choose a theme.
✓We can choose the theme from one of the following –
Environment Agriculture Healthcare Infrastructure Education
women safety Transport Wellness Cyber security etc.
✓Theme can also be selected from 17 SDGs
Sustainable Development Goals (SDGs) are a collection
17 global goals ,
169 Targets
set by United Nations in 2015
(and to be implemented by 2030.

14. Steps of Problem Scoping:
✓Choose a Theme and Topic of AI Projet.
✓Identify the Problems around the selected Topic
✓Setting Goal for an AI Project

15. AI Project Cycle

16. ✓Whenever we are starting any work, certain problems
always associated with the work or process.
✓These problems can be small or big
✓sometimes we ignore them, and sometimes we need
urgent solutions.
✓Problem scoping is the process by which we figure
out the problem that we need to solve.
Problem Scoping:

17. The 4Ws Canvas

19. STAKEHOLDER
+In project management, a stakeholder is
anindividual or a group of people who are
impacted by the outcome of a project/product.
They are interested in the success of the product
and can be within or outside the company that is
sponsoring the project.

20. The 4Ws Canvas -The 4Ws Canvas is a helpful tool in Problem Scoping.
1. Who?:
•who is facing a problem,
•who are the stakeholders of the problem,
•who are affected because of the problem.
2. What?:
✓what the problem
✓what you know about the problem.
✓What is the nature of the problem?
✓Can it be explained simply?
✓ How do you know it’s a problem?
✓What is the evidence to support that the
problem exists?
✓What solutions are possible in this
situation? etc.
At this stage, you need to determine the exact
nature of the problem.
3. Where?
✓It is related to the
✓Context or
✓situation or
✓location of the problem
4. Why?:
✓The reason we need to solve the problem.
✓The benefits which the stakeholders would
get from the solution and
✓How would it benefit them as well as
society?
✓what are the benefits to the Stakeholders
after solving the problem?

21. PROBLEM STATEMENT
A Problem Statement is a short, Clear description in words of
the problem listing its stakeholders, their issues, context and
reasons to solve the problem. i.e.
✓Once the 4W Problem canvas has been completed, all the four
cards need to be summarized into one template called the
problem statement.
✓The Main Goal of making a Problem Statement
template is to convert a generalized problem into a
well-defined problem.

22. DATA
ACQUISITION
✓Where data required for the project is acquired in
specific forms and formats.
✓In Data Acquisition, data is understood, gathered,
filtered, cleaned, and finally stored in a data
storage system.

23. Project Cycle #easy2learn with jyoti
Data Acquisition consists of two words:
Data: Data refers to the raw facts, figures, or pieces of facts, or
statistics collected for reference or analysis.
Acquisition: Acquisition refers to acquiring data for the
project. The stage of acquiring data from relevant sources is known
as data acquisition.

24. TYPES OF DATA USED IN
AI PROJECT
Project Cycle #easy2learn with jyoti
1. Structured Data
2. Unstructured Data

25. 1. Structured Data: 2. Unstructured Data:
•Data in specific forms such
as tabular form.
•Doesn’t have any specific pattern or constraints as well
as can be stored in any form.
•Not have any predefined data Model.
•Mostly the data that exists in the world is unstructured
data.
Example~
•The cricket scoreboard,
•Your school timetable,
•Exam datasheet, etc.
Examples~
•Youtube Videos,
•Facebook Photos,
•In an AI system for analyzing the most popular social
media posts, the data – Social-media-post does not
have a predefined structure it can be text or video or a
link or an image .
STRUCTURAL CLASSIFICATION

26. •Both Structured and unstructured Data have certain data features.
•Data feature refers to the type of data we want to collect for the AI
system.
•Example –
• AI system used for predicting average saving, the data feature could be – salary/
income, average spending, average saving, etc.
•Similarly, for an AI System analyzing social media posts, the data features
required would be – social-media –posts, platform and time-posted etc.
DATA FEATURE- refers to the type of data, having a name, collected
for a project.

27. Quality Data Characteristics
•Accuracy – is the data accurate as per timeliness & real data?
•Relevance- is it relevant to the intended purpose & is it really needed?
•Reliability – is the data collected from reliable and authentic sources?
•Timeliness –Has the data been collected quickly after the event? Is it up to
date?
•Validity – is the data match with the requirement.
•Completeness- has invalid & incomplete data been removed?
AI Projcet Cycle # easy2learn with jyoti

28. Reliable data source
AI Projcet Cycle # easy2learn with jyoti

29. 1. Interview ~~ one-to-one conversation
2. Survey ~ A survey refers to a study of the opinions, responses, etc. of a group of stakeholders
3. Observations – watching / Noticing/perceiving of what people do or what events take place in a specific working
environment.
4. Application Programming Interface ~ works behind a popular software program or game to collect specific types of
data. E.g. social media programs interface, data like people’s most preferred game, most liked post, most used time, etc.
may be gathered.
5. Web Scrapping – Web Scrapping refers to a data collection technique using a tool called Web Scrapper that extracts
data from a website.
6. . Sensors ~ - sensors or electronic sensors can measure various parameters such as weather, humidity, body
temperatures, blood pressure, heartbeat, weight, and many more. (“Apple Watch”, Fitbit makes good use of sensors.
Fact- IoT can’t function without sensors).
7. Cameras - collect data graphically/video
8. Problem Reports - Authentic documents listing the problem of a system either after conducting some tests.
9. The Internet - take data from the Internet only after ensuring –
The source of the data is authentic and Licensed for public use through license. We can take from some open-source websites
hosted by the govt. such as data.gov.in, india.gov.in, censusindia.gov.in, etc.
AI Projcet Cycle # easy2learn with jyoti

30. •1. Primary Data - is the type that we gather by ourselves. It means we are actively
involved in the sourcing of information.
•Secondary Data – is all around us. It is easily accessible on the Internet.
•The Collection of Primary data has been done by someone else before getting
uploaded to the internet, Secondary data comes in the form of search results.
AI Projcet Cycle # easy2learn with jyoti

31. DATA EXPLORATION WITH
DATA VISUALIZATION
AI PROJECT CYCLE
STAGE 3
AI Projcet Cycle # easy2learn with jyoti
ARTIFICIAL INTELLIGENCE - 417
UNDERSTANDS THE
NATURE OF DATA.

32. DATA EXPLORATION
WITH DATA VISUALISATION
UNDERSTANDS THE NATURE OF DATA.
AI Projcet Cycle # easy2learn with jyoti

33. AI Projcet Cycle # easy2learn with jyoti
During Data Acquisition, the data, which is gathered from
various sources, is often in large, unstructured volumes. Thus, it
is first cleaned and handled to bring it in a form useful for data
analysis & then data exploration techniques.
Need for Data Exploration
After Data Acquisition, the data needs to be cleaned by removing
redundant or unrequired data and handling missing values &
then its characteristics are to be thoroughly understood. All this
process is known as DATA EXPLORATION.
OR

34. AI Projcet Cycle # easy2learn with jyoti
What is DATA EXPLORATION?
DATA EXPLORATION uses various techniques, such as
✓DATA VISUALIZATION AND
✓STATISTICAL TECHNIQUES
to describe dataset characterization in order to better understand the
nature of the data by showing the trends and patterns in data.

35. DATA EXPLORATION TECHNIQUES ARE APPLIED MAINLY :
✓To visually explore & identify relationships between different
data variables.
✓To understand the Structure of the dataset.
✓To identify the presence of data points that differ significantly
from other observations (called outliers)
✓To obtain the distribution of data values in order to reveal trends
& patterns & points of interest.
AI Projcet Cycle # easy2learn with jyoti

36. DATAVISUALISATION
•Data Visualization Is the process of representing data
visually or graphically
• By using visual elements like
Charts
Graphs
Diagrams
Maps etc.
AI Projcet Cycle # easy2learn with jyoti

37. Importance of Data Visualisation
powerful way to represent a bulk of data in a collective visual form.
Explore data with presentable results.
Makes it easy to interpret & comprehend data.
Makes it easier to see the relationships and trends of data.
Combines categories of data and thereby reducing data for processing.
Helps in defining strategy for using data for AI model to be developed at a
later stage.
AI Projcet Cycle # easy2learn with jyoti

38. WAYS TO VISUALIZE DATA
TechniqueFormat Example Description
Charts Scatterplot, Bubble Chart
[Used with Numeric Type of
Data]
Scatter Plot – XY candidates
Bubble Chart – X(Independent
Variable) Y(dependent Variable)
marked in Bubble form. Y values
determine the Bubble size. A bigger
marker means a bigger value.

AI Projcet Cycle # easy2learn with jyoti

39. GraphsLine graph, pie graph, bar graph,
histogram
[Used with Numeric Type of Data]
Histogram – Unlike a bar graph,
there are no gaps in between in a
histogram.
A histogram is used to summarize
discrete or continuous data.

AI Projcet Cycle # easy2learn with jyoti

40. Maps [Used with processed numeric data
linked with textual units]
Choropleth – used with statistical
data (Numeric, processed data)
attached to enumeration(
गणना) units (textual data e.g.
countries, provinces(प्ाांतों), states,
etc. e.g.-
•World map of Covid 19 spread,
country-wise.
•Map showing the percentage
increase in real estate value,
state-wise.
•World map of income tax rates,
country-wise.
,

Heat map depicted through color codes.
•A Geographical heatmap of high & low
population/network density displaying
points on a map through different colors.
•A Stock index heatmap depicting
prevailing trends in the market through
colors.
• AI Projcet Cycle # easy2learn with jyoti

41. Temporal Timeline
{Shows a series of Events
in Chronological order.}
Network Node-link diagram
{Shows how things are
interconnected through
the use of nodes /vertices
& link lines to represent
their connection}
AI Projcet Cycle # easy2learn with jyoti

42. Word Cloud
•The word cloud visualization technique represents the
frequency of a word within a body of text with its relative size
in the cloud.
•BIG DATA
AI Projcet Cycle # easy2learn with jyoti

44. Scatter Plot If you have two variables that pair well together, plotting them on a scatter
diagram is a great way to view their relationship and see if it’s a positive or negative
correlation.
Bubble Chart An extension of a scatterplot, abubble chartis commonlyusedto visualize
relationships between three or more numeric variables.
Bubble charts are often usedto present financial data.
Line Chart to track changes over short and long periods of time
Pie Chart Apie chartexpresses a part-to-whole relationship in your data.

46. Duration: 75min
MODELLING
Grade: 4 CCSS,NGSS
Modelling is the phase during which the AI
model for the desired outcome is trained
using the collected data repeatedly until it
starts producing desired results.
AI Projcet Cycle # easy2learn with jyoti

47. ✓In Modelling Data is represented mathematically.
✓The representation of data in mathematical
equations & other forms is necessary for
developing AI models.
✓The ability to mathematically describe the
relationships between data parameters forms the
core of every AI model.
AI Projcet Cycle # easy2learn with jyoti

48. Categories of AI ModelRule Based AI Model Learning Based AI Model
AI Projcet Cycle # easy2learn with jyoti

49. RULE-BASED AI MODEL
Thus, the machine would follow the
algorithm’s rules or instructions & perform the
given tasks or take the decisions accordingly.
In this, an AI system is developed using the predefined
labels,
rules,
patterns, or
relationships
as given by the developers in the algorithm.
AI Projcet Cycle # easy2learn with jyoti

50. Drawbacks of Rule-Based Model
Lot of Manual Work –As all the rules governing the decisions
must be pre-coded & made available to the system.
Consumes a lot of time –creating all possible rules for a
system requires a lot of time. The bigger & more complex the system is,
the more time-consuming it becomes.
Suitable only for less complex domains-Covering
all the combinations & permutations of rules for a complex system is
challenging for a rule-based system.
Limited adaptability & learning capacity –Rule-
based systems rules are predefined & do not get updated on their own.
Static & not scalable –Can’t update their own, so these
systems are static & not scalable.
AI Projcet Cycle # easy2learn with jyoti

51. FACTS
Rule Based models are often preferred
for limited-scale projects that require
✓ limited efforts,
cost,&
updates.
In other words, the rule-based systems
are not largely scalable.
AI Projcet Cycle # easy2learn with jyoti

52. Also known as the data-driven AI model.
In this a lot of data is shown & questions/answers are asked
in order totrain the system about the right answer.
For example –to train a system about recognizing cats, it
would be shown lots & lots of images of cats, & other animals
& letting it know when it “guessed” it correctly or not.
After millions of training cycles, it will “learn” to get it
increasingly right.
This is used when data is unknown or random or unlabelled.
Both Machine Learning (ML) & Deep Learning (DL) fall under “Learning Based AI.”
AI Projcet Cycle # easy2learn with jyoti

53. Machine Learning
it enables machines to automatically
learn & improve at tasks with
experience & by the use of data.
ML-based machines undergo lots of
repetitions of taking data & testing it;
these then keep track of when things
went wrong or right & keep
improving their results.
Deep Learning
Deep Learning (DL) is a subset of ML
where learning takes place through
examples.
it filters the input data using layers &
rule-based algorithms to predict &
classify information.
Tasks like Speech & Image recognition
are performed through a deep learning
system.
A driverless car used these technologies.

AI Projcet Cycle # easy2learn with jyoti

54. Labelled Data
It is a group of samples that have been
marked with one or more labels.
Labeling puts meaningful tags to data
so that it gives some information or
explanation about the data.
For example –
1.if some audio clips are labeled as
“SPEECHES”. Then they certainly
give information that these contain
speeches of some people.
2.If an article is labeled as a “news
article” it gives information that
it contains some sort of news.

unlabelledData
It is a description of data that have not
been tagged with labels identifying
characteristics, properties or
classification of data.
there is no explanation for each piece of
unlabelled data – it just contains data and
nothing else.
Examples – photos, audio, recording,
videos, news articles, tweets, etc.

55. The learning-based models are
developed through three different
approaches
Supervised Learning
Unsupervised Learning
Reinforcement Learning
AI Projcet Cycle # easy2learn with jyoti

56. Supervised Learning
✓Supervised Learning is a machine learning approach in which a
machine, with the help of an algorithm (called the model), learns on
labelled datasets & is later tested with some unlabelled data whose
answers are pre-known to evaluate its accuracy on training data.
✓For example, labelled datasets of flower images would contain photos of
roses tagged as roses, photos of daisies tagged as daisies, and so on for
other flowers.
When shown a new image, the model compares it to the
training examples to predict the correct label.
The model gets feedback about its results as per the desired
outputs & this way, it learns to classify correctly.

AI Projcet Cycle # easy2learn with jyoti

57. Supervised Learning
AI Projcet Cycle # easy2learn with jyoti

58. DISCRETE DATA CONTINUOUS DATA
DISCRETE DATA IS COUNTABLE Opposite of discrete data i.e it is
measurable.
IT INVOLVES INTEGERS ONLY Between 1 and 2 lie other numbers
such as 1.5, 1.7,1.9, etc.
EXAMPLE –
1. NUMBER OF PLAYER IN A TEAM.( COUNTABLE
ONLY – YOU CAN’T SAY THE TEAM HAS 11.5
PLAYER)
2. NUMBER OF STUDENTS IN A
CLASS. NUMBER OF COMPUTER IN A LAB
3. NUMBER OF QUESTIONS IN A EXAM
Example
Height, weight, temperature, time,
and so on.
AI Projcet Cycle # easy2learn with jyoti

59. Types of
Supervised Learning
Based on discrete data or
continuous data Supervised
Learning is mainly divided into
two categories—
AI Projcet Cycle # easy2learn with jyoti

60. The Classification models use non-continuous i.e. discrete data. The Classification
algorithm is used to identify the category of new observations based on training data.
The best example of an
✓The ML classification algorithm isan Email Spam Detector.
✓Is it going to rain or not?
✓Is this a picture of a specific animal
✓Is this social media post negative or positive (predict the sentiment)
In the diagram, there are two classes, class A and Class B.
These classes have features that are similar to each other and
dissimilar to other classes.
AI Projcet Cycle # easy2learn with jyoti
X

61. The Regression models use a continuous data set. The Regression AI model is based
on a mathematical approach used to find the relationship between two or more
variables & predict outcomes.
The best example of an
1.What would be the temperature of the city tomorrow if factors like historical
temperature data, predictions, wind, humidity, etc. are available?
2.What would be the house prices in the wake of factors like square footage,
location & proximity to schools, hospitals, public transport, etc.?
3.What all colleges & streams would be within reach if the qualifying scores
are known?
4.How would the prices of a specific fruit be affected if its production is
increased and there is the overall dip in production cost can easily be
determined through a regression model.
AI Projcet Cycle # easy2learn with jyoti

62. Advantages & Disadvantages of Supervised Learning
Advantage
1.Computationally less complex.
2.Highly accurate & trustworthy methods.
3.Very useful in cases when a user has exact data about the class of projects.
Disadvantage
1.Computation time is very high.
2.Since it works with labeled data, the data must be pre-processed to be in the correct form.
Pre-processing of data has a huge impact on the overall training of the system.
3.The Training data must be based on real good working examples as their absence would
largely affect the efficiency of the data.
4.Supervised learning system requires continuous updating with all the new learning &b
findings.
AI Projcet Cycle # easy2learn with jyoti

63. Some Real-world Application of Supervised Learning
1.Biometric Identification. Biological Information of humans such as fingerprints, iris texture, earlobe, and
so on can be stored electronically in terms of some patterns. Modern devices are trained in a supervised
way to recognize and identify people based on their biometric information, e.g., fingerprint unlocking or
facial recognition by cell phones or other devices.
2.Speech Recognition. Machines and devices of today can be taught in a supervised way to recognize how
you speak. Using this, the machine can recognize your voice through tonal quality, voice throw, and
diction. The most common examples of speech recognition are virtual assistants such as Google Assistant,
Alexa, and Siri.*
3. Spam Detection. Algorithms can be trained to identify the emails with specific keywords to be termed as
spam, e.g., "Congratulations on winning so and so" and so forth. These days even apps are available to
which we can choose to tell which keywords need to be blocked and the app will block those messages
having the keyword.
4.Object-Recognition for Vision. Under supervision, the machine can be trained to identify something and recall
the Emoji Scavenger Hunt game you played. For this, you teach your algorithm with a set of data and their
predicted result. Using this, the machines or devices learn to identify and recognize a new instance.
AI Projcet Cycle # easy2learn with jyoti

64. Unsupervised Learning
In Unsupervised learning, as the name suggests, there is no supervision, no feedback, no
pre-known/ desired outputs & not even any labeled data. Thus Unsupervised Learning is a
learning approach of the machine where the machines with the help of an algorithm (the
model) learn on an unlabelled dataset where it categories data on the basis of common
characteristics, features & patterns.
In Unsupervised Learning Based AI Model ( the algorithm) finds the patterns, trends, and
features, and clubs the data having the same patterns in one group ( or cluster).
For every new input (also unlabelled), it tries to put it in a cluster as per its pattern or
characteristics and then enables new data to be categorized into an existing cluster
AI Projcet Cycle # easy2learn with jyoti

65. AI Projcet Cycle # easy2learn with jyoti
Unsupervised Learning

66. ✓Thus unsupervised learning technique works with unlabelled
data & creates clusters of items having similar features,
characteristics or patterns.
✓The training dataset of an unsupervised learning-based AI
model, is a collection of unlabelled data without a specific
desired outcome or correct answer. The AI model then
attempts to automatically find structure or patterns in the
data by extracting useful features & analysing its structure.
✓For example, after giving it a set of images of animals (as
training data) without any label/tag/explanation, it would try
to combine images having similar features or characteristics
in one cluster.
AI Projcet Cycle # easy2learn with jyoti

67. ✓Unlike supervised learning where the training
data set contains labeled data with the
corresponding outputs, unsupervised–learning
works with unlabelled training data without any
corresponding outputs.
AI Projcet Cycle # easy2learn with jyoti

68. TYPES OF UNSUPERVISED LEARNING
Clustering
Association
Dimensionality
Reduction
AI Projcet Cycle # easy2learn with jyoti

69. CLUSTERING
AI Projcet Cycle # easy2learn with jyoti
It groups unlabelled data based on their similarities or
differences.
An AI model without even being an expert ornithologist (bird specialist),
can take a collection of bird photos and separate them roughly by
species, relying on cues like feather color, size, and beak shape. etc.

70. Example – Some examples of Clustering problems/applications are –
1.PATTERN RECOGNITION – A group of Cancer Patients may be considered for a
special type of treatment on the basis of their gene expression measurement.
2.IDENTIFYING FAKE NEWS – by clustering articles with high percentage of
sensationalizing & click-bait terms.
3.Document Analysis – by clustering & organizing similar documents quickly using
the characteristics identified in the paragraphs of multiple documents.
AI Projcet Cycle # easy2learn with jyoti

71. Example 1: Retail Marketing
Retail companies often use clustering to identify groups of households that are similar to
each other.
For example, a retail company may collect the following information on households:
•Household income
•Household size
•Head of household Occupation
•Distance from nearest urban area
They can then feed these variables into a clustering algorithm to perhaps identify the
following clusters:
•Cluster 1: Small family, high spenders
•Cluster 2: Larger family, high spenders
•Cluster 3: Small family, low spenders
•Cluster 4: Large family, low spenders
•The company can then send personalized advertisements or sales letters to each
household based on how likely they are to respond to specific types of advertisements.
AI Projcet Cycle # easy2learn with jyoti

72. Example 2: Streaming Services
Streaming services often use clustering analysis to identify viewers who have
similar behavior.
For example, a streaming service may collect the following data about individuals:
•Minutes watched per day
•Total viewing sessions per week
•Number of unique shows viewed per month
Using these metrics, a streaming service can perform cluster analysis to
identify high usage and low usage users
so that they can know who they should spend most of their advertising dollars on.
AI Projcet Cycle # easy2learn with jyoti

73. Example 5: Health Insurance
Actuaries(clerks) at health insurance companies often used cluster analysis to
identify “clusters” of consumers that use their health insurance in specific ways.
For example, an actuary may collect the following information about
households:
•Total number of doctor visits per year
•Total household size
•Total number of chronic conditions per household
•Average age of household members
An actuary can then feed these variables into a clustering algorithm to identify
households that are similar.
The health insurance company can then set monthly premiums based on how
often they expect households in specific clusters to use their insurance.
AI Projcet Cycle # easy2learn with jyoti

74. Clustering algorithms use both supervised and
unsupervised learning methods.
AI Projcet Cycle # easy2learn with jyoti

75. DIMENSIONALITY REDUCTION
➔Dimensionality Reduction broadly means representing an object in smaller
dimensions.
➔The number of input features, variables, or columns present in a given dataset
is known as dimensionality, and the process to reduce these features is called
dimensionality reduction.
➔It is a way of converting the higher dimensions dataset into a lesser
dimensions dataset ensuring in a way so that the actual meaning & intent of
the original dataset is not lost.
➔These techniques are widely used inunsupervised learningapproaches.
AI Projcet Cycle # easy2learn with jyoti

76. Some examples of dimensionality reduction applications are –
➔Video & Satellite observation compression
➔Email Classification
➔speech recognition,
➔signal processing,
➔ BIOINFORMATICS
➔ data visualization,
➔noise reduction,
➔cluster analysis, etc.
AI Projcet Cycle # easy2learn with jyoti

77. ASSOCIATION
The association is an unsupervised learning technique that finds important relations
between variables or features in a data set.
For example – As Recommendation Systems
1.if you pick some home décor items such as lamps or shelves in an online shopping
cart, it will start suggesting related items such as furniture, rugs & even interior
designing firms.
2. People that buy a new home are most likely to buy new furniture & thus suggest
furniture items & stores to them.
1.As Anomaly Detection – a crit card is us
For example, if a credit card is used at the same time in two distant cities it points to
fraud or anomaly.
AI Projcet Cycle # easy2learn with jyoti

78. 1.Useful in finding unknown patterns & features in data.
2.Very useful in many real-world situations where labeling of data is not
feasible.
3.Training of Model takes place in real-time.
Advantage of
unsupervised Learning
AI Projcet Cycle # easy2learn with jyoti

79. 1.Not easy to comment about the accuracy of the results as input datasets in
unlabelled & its structure is largely unknown. After a long training, the
results can be thought of moving towards accuracy.
2.The user needs to spend time interpreting & labeling the classes generated
by the system after training.
Disadvantage of
unsupervised Learning
AI Projcet Cycle # easy2learn with jyoti

80. It refers to the learning approach where the training dataset has both labelled &
unlabelled data.
Using labelled data supervised learning techniques are used & using
unsupervised learning techniques new unexplored features are extracted from
the unlabelled data.
Semi-supervised learning is especially useful for medical images like CT scans or
MRIs.
Semi-Supervised Learning
AI Projcet Cycle # easy2learn with jyoti

81. In the reinforcement Learning approach, the AI Model (the algorithm, also called
the agent) iteratively attempts to accomplish a particular goal, or improve
performance on a specific task, in the best possible way known to it. If the
action/step of the agent is helpful towards achieving the goal, it is given a
reward. The overall aim of the agent is to predict the best next step to take to
earn the biggest final reward.
This technique is especially useful for training robots, which make a series of
decisions in tasks such as walking through dangerous situations like
building on fire avoiding the places of fire,
steering an autonomous vehicle, or
managing inventory in a warehouse.
Reinforcement Learning
AI Projcet Cycle # easy2learn with jyoti

82. There are two types of Reinforcement Learning—
1.Positive Reinforcement – focuses on increasing rewards to encourage certain
behavior.
2.Negative Reinforcement – Focuses on removing to encourage a certain type of
behavior.
Types Reinforcement Learning
AI Projcet Cycle # easy2learn with jyoti

83. Positive Reinforcement Learning Example
Positive behavior followed by positive consequences
Managers rewards
the employee for
productivity
Teacher praises the
Student for effort
Student is more
active in class
Employee Increases
her/his productivity
AI Projcet Cycle # easy2learn with jyoti

84. Negative Reinforcement Learning Example
Positive behavior followed by removal of negative consequences
Teacher stops
criticizing students’s
input
Student is more likely
to participate in
discussion.
Employee stops
nagging (नैगगांग
-गिड़गिडा)the
employee
नैगगांग
Employee Employee
starts being more
productive
AI Projcet Cycle # easy2learn with jyoti

85. Practical Application of Reinforcement Learning Example
✓Training Self-Driven Cars about how to drive on roads, in traffic situations, with speed limits. In
case of any stationary or moving obstacles in front of the car and so on.
✓Training robots to work in hazardous situations.
✓Training robots to perform difficult & dangerous tasks in various industries.
✓Text summarizing engines & Dialogue agents that use text and speech(NLP) such as SIRI, Alexa,
Google Assistant, etc.
✓In Healthcare making time-dependent decisions.
✓In gaming learning ways to better & better gameplay & strategies.
AI Projcet Cycle # easy2learn with jyoti

86. Unsupervised Learning vs. Supervised Learning
PropertiesSupervised Learning Unsupervised Learning
Definition Supervised Learning is the type of machine
learning that happens under human
supervision
Unsupervised Learning is the type of machine
learning that happens without human
supervision. A machine tries to find any
pattern in data by itself.
Applicable inClassification and Regression problemsClustering & Association problems.
Accuracy of the
results
Provide more accurate results May provide less accurate results.
Input Data Labelled Unlabelled
Use cases Spam filters
Demand forecasting
Price Prediction
Image recognition
Recommender systems
Anomaly detection
Customer segmentation

87. THANK YOU