MA
Artificial Intelligence &
Machine Learning Overview
By Mukesh Sinha
AGENDA
ARTIFICIAL INTELLIGENCE – INDUSTRY OVERVIEW
MACHINE LEARNING – OVERVIEW
TYPES OF MACHINE LEARNING
MACHINE LEARNING – PROBLEM CLASSIFICATION
CASE STUDIES & DEMO
ARTIFICIAL INTELLIGENCE – INDUSTRY OVERVIEW
Investment in AI is growing at high rate
In 2016 , Companies invested
$26B to $39B
in artificial intelligence
TECH GIANTS
$20B to $30B
STARTUP
$6B to $9B
3x External Investment growth since 2013
AI Adoptions – Top 3 industries spending ( % of Spend)
Banking
20%
of Spend
Intelligent Financial
health assessment
Credit scoring worthiness
Healthcare
18%
of Spend
Rapid diagnostic with
Medical images
X-Ray, MRI etc
Retail
17%
of Spend
Interactive personalized
experience analyzes
Shoppers History , mood
, expressions etc
Data around the World Top 5 AI Companies
Face Recommendation Navigation Product Recommendation Speech Recognition Medical System
Benefits
Digitially Mature
Larger Business
• JP Morgan has been able to save 360,000 work hours by automating part of their lawyer work by leveraging an
operation efficacy applications
• UBS implemented AI based platform for post-trade allocation requests . Investment banker took about 45 minutes,
to complete the end to end process , AI implementation helped banker to do task in 2 minutes only
• Korean airlines built AI based expert system to redcue the number of flight dealy and cancellations and shortened its
mainteanance lead times by 90%
• NetFlix leaverages AI to tag its movie content and built recommender system for improving customer experience by
offering personalized product recommendations and helping customers to find desired product faster
Increased Automation Productivity
Uncovering new insights
Increased Employee Productivity
• We are pursuing AI to empower every person and every institution ..so that they can go on to solve the most
pressing problems of our society and economy - Satya Nadela , Microsoft
• Building genral artificial intelligence (AI) in a way that helps people meaningfully – I think the word moonshot
is an understatement for that . - Sunder Pichai , Google
• We are living in the golden age of AI , We are now solving problems with machine learning and artificial
intelligence - Jeff Bezos Amazon
• Artificial Intelligence is the new electricity. - Andrew Ng , Baidu
Top Industry Leader Statement on AIAI Use Cases
90 % of World‘s
Data created in past
2 years
3 Billion Online
In 2000, only 738 million
used internet, In 2015
number grew to 3.2 Billion
According to World
Bank, 75% of people
own a cell phone
204 Million
emails are
sent each minute
MACHINE LEARNING OVERVIEW
Machine Learning Used
Machine Learning Tools and Services
Trend Forecasting Fraud Detection Price Prediction Product Recommendation
Interesting Case Studies
Large US based Retail Grocery Store
Observation : Men between 30-40 years in age shopping between 5
PM and 7 PM on Fridays, who purchased diapers along with beer in
their shopping cart
Analysis : Data Analysis revealed that new fathers tend to buy more
beer , because they are spending less time at the pub.
Solution: Used Data Mining and Anaytics approach Retailer shop
move the beer isle closer to diaper isle
Benefits :35 % increase in sales of both
Large Global Insurance company - AXA
Use Case : Approximately 7-10% of AXA’s customers cause a car
accident every year, about 1% are so-called large-loss cases that
require payouts over $10,000.
Objective : AXA adjusters to understand which clients are at higher
risk in order to optimize the pricing of its policies.
Solution : AXA’s R&D team analyzed the historical data ( Age of
Driver , Address , Age of car etc) and using Machine learning
algorithm to optimizing price by large-loss traffic accident
Benefits : Team achieved 78% accuracy in its predictions
Significant advantage for optimizing insurance cost and pricing,
Machine Learning - How it works
Learn from Experience Learn from Data Follow Instructions
What is Machine Learning ?
Machine learning is an application of artificial
intelligence (AI) that provides systems the
ability to automatically learn and improve from
experience without being explicitly
programmed.
Human ComputerMachine Learning
TYPES OF MACHINE LEARNING
• Make machine learn explicitly
• Data with clearly defined output is given
• Direct feedback is given
• Predicts outcome or future
Supervised Learning
Teacher teaches kids and
providing the solution to them
Unsupervised Learning
• Machine understand the data ( identifies
pattern and data
• Evaluation is qualitative or indirect
• Does not predict or find anything specific
Predicting the weather forecast
Kids are taking decisions out of
their own understanding
Reinforcement Learning
• Reward based learning
• Learning from +ve and –ve reinforcement
• Machine learns how to act in certain environment
• To Maximize rewards
Kids will take actions on his own
from past exp, and parent
providing the feedback
Decision Tree Random Forest Naïve Bayes Support Vector machineK-Means ClusteringTensorFlow
MACHINE LEARNING ALGORITHM
You Tube Recommendation videos System which play chess
Linear Regression Logistic Regression
MACHINE LEARNING – PROBLEM CLASSIFICATION
Classification Algorithm
Anomaly Detection Algorithm
Regression Algorithm
Clustering Algorithm
Reinforcement Algorithm
• Used to classify record
• Limited number of answers
• Analyze the pattern
• Alerts in case of change in pattern
• Used to calculate numeric value
• Predicts outcome
• Sepeartes the data into groups or clusters
• Ease out the interpreation of data
• Design how brains of humans respond
• Learn from outcome , reward , punishment
Email Spam for Gmail and Yahoo
Prediction of house price
QUESTIONS ML ALGORITHM ALGORITHM DESCRIPTION EXAMPLE
Fraud Detection for Credit card
Product Recommendation
Google self drive car
How much or how many ?3
Is this weird ?2
Is this A or B?1
How is this organized ?4
What should I do next?5
Business Problem
ML DEMO - PORTUGUESE BANK
age job marital education default balance Housing loan contact day month duration campaign pdays previous poutcome Y
30 unemployed married primary no 1787 no no cellular 19 oct 79 1 -1 0 unknown No
33 services married secondary no 4789 yes yes cellular 11 may 220 1 339 4 failure no
30 management married tertiary no 1221 yes no telephone 25 jul 279 4 -1 0 unknown yes
• There has been a revenue decline for the Portuguese bank and they would like to know what actions to take.
• After investigation, they found out that the root cause is that their clients are not depositing as frequently as before.
• Knowing that term deposits allow banks to hold onto a deposit for a specific amount of time, so banks can invest in higher gain financial products to make a profit.
• In addition, banks also hold better chance to persuade term deposit clients into buying other products such as funds or insurance to further increase their revenues.
• As a result, the Portuguese bank would like to identify existing clients that have higher chance to subscribe for a term deposit and focus marketing effort on such clients.
To predict which clients are more likely to subscribe for term deposits
Data Set Description
Name Description and Values
Personal Client Information
Age Age at the contact date (Numeric ≥18)
Marital Status Married, Single , Divorced, Widowed etc
Sex Male or Female
Job Unemployed , 'entrepreneur, technician
Name Description and Values
Bank Client Information
Annual Balance in Euro currency (Numeric )
Default ( Credit) Yes , No , Unknown
Housing Loan Yes , No , Unknown
Pesonal Loan Yes , No , Unknown
Name Description and Values
Last contact information
Contact Type contact communication type
Date When the contact was made
Duration Duration of the contact
Name Description and Values
Other Attributes
campaign No of contacts performed during this campaign
pdays No of contacts performed before this campaign
poutcome Outcome of the previous compaign
Output variable (desired target):
y - has the client subscribed a term deposit? (binary: 'yes','no')
Data Set
Business Goal
PYTHON DEMO STEPS FOR BANKING SCENARIO
Import Python Library Packages Pandas , numpy, matplotlib
Loading the Banking data and reading the file
Drop the variables which are not required for prediction
Spliting the data into training set and testing set
Applied logistic regression model using training data set
Checking the score or accuracy of the model
Library File
Data Set
Splitting the Data Set
Data Filter
ML Model
Accuracy
DESCRIPTION PYTHON CODESTEPS
REFERENCES
Machine Learning book - Tom Mitchel The Elements of Statistical Learning book -Trevor Hastie
Video Lecture - Andrew NG
Video Lecture - Yaser Abu Mostafa
Artificial Intellignece Reports from IDC , PWC , DB and McKinsey

Artificial Intelligence and Machine learning overview

  • 1.
    MA Artificial Intelligence & MachineLearning Overview By Mukesh Sinha
  • 2.
    AGENDA ARTIFICIAL INTELLIGENCE –INDUSTRY OVERVIEW MACHINE LEARNING – OVERVIEW TYPES OF MACHINE LEARNING MACHINE LEARNING – PROBLEM CLASSIFICATION CASE STUDIES & DEMO
  • 3.
    ARTIFICIAL INTELLIGENCE –INDUSTRY OVERVIEW Investment in AI is growing at high rate In 2016 , Companies invested $26B to $39B in artificial intelligence TECH GIANTS $20B to $30B STARTUP $6B to $9B 3x External Investment growth since 2013 AI Adoptions – Top 3 industries spending ( % of Spend) Banking 20% of Spend Intelligent Financial health assessment Credit scoring worthiness Healthcare 18% of Spend Rapid diagnostic with Medical images X-Ray, MRI etc Retail 17% of Spend Interactive personalized experience analyzes Shoppers History , mood , expressions etc Data around the World Top 5 AI Companies Face Recommendation Navigation Product Recommendation Speech Recognition Medical System Benefits Digitially Mature Larger Business • JP Morgan has been able to save 360,000 work hours by automating part of their lawyer work by leveraging an operation efficacy applications • UBS implemented AI based platform for post-trade allocation requests . Investment banker took about 45 minutes, to complete the end to end process , AI implementation helped banker to do task in 2 minutes only • Korean airlines built AI based expert system to redcue the number of flight dealy and cancellations and shortened its mainteanance lead times by 90% • NetFlix leaverages AI to tag its movie content and built recommender system for improving customer experience by offering personalized product recommendations and helping customers to find desired product faster Increased Automation Productivity Uncovering new insights Increased Employee Productivity • We are pursuing AI to empower every person and every institution ..so that they can go on to solve the most pressing problems of our society and economy - Satya Nadela , Microsoft • Building genral artificial intelligence (AI) in a way that helps people meaningfully – I think the word moonshot is an understatement for that . - Sunder Pichai , Google • We are living in the golden age of AI , We are now solving problems with machine learning and artificial intelligence - Jeff Bezos Amazon • Artificial Intelligence is the new electricity. - Andrew Ng , Baidu Top Industry Leader Statement on AIAI Use Cases 90 % of World‘s Data created in past 2 years 3 Billion Online In 2000, only 738 million used internet, In 2015 number grew to 3.2 Billion According to World Bank, 75% of people own a cell phone 204 Million emails are sent each minute
  • 4.
    MACHINE LEARNING OVERVIEW MachineLearning Used Machine Learning Tools and Services Trend Forecasting Fraud Detection Price Prediction Product Recommendation Interesting Case Studies Large US based Retail Grocery Store Observation : Men between 30-40 years in age shopping between 5 PM and 7 PM on Fridays, who purchased diapers along with beer in their shopping cart Analysis : Data Analysis revealed that new fathers tend to buy more beer , because they are spending less time at the pub. Solution: Used Data Mining and Anaytics approach Retailer shop move the beer isle closer to diaper isle Benefits :35 % increase in sales of both Large Global Insurance company - AXA Use Case : Approximately 7-10% of AXA’s customers cause a car accident every year, about 1% are so-called large-loss cases that require payouts over $10,000. Objective : AXA adjusters to understand which clients are at higher risk in order to optimize the pricing of its policies. Solution : AXA’s R&D team analyzed the historical data ( Age of Driver , Address , Age of car etc) and using Machine learning algorithm to optimizing price by large-loss traffic accident Benefits : Team achieved 78% accuracy in its predictions Significant advantage for optimizing insurance cost and pricing, Machine Learning - How it works Learn from Experience Learn from Data Follow Instructions What is Machine Learning ? Machine learning is an application of artificial intelligence (AI) that provides systems the ability to automatically learn and improve from experience without being explicitly programmed. Human ComputerMachine Learning
  • 5.
    TYPES OF MACHINELEARNING • Make machine learn explicitly • Data with clearly defined output is given • Direct feedback is given • Predicts outcome or future Supervised Learning Teacher teaches kids and providing the solution to them Unsupervised Learning • Machine understand the data ( identifies pattern and data • Evaluation is qualitative or indirect • Does not predict or find anything specific Predicting the weather forecast Kids are taking decisions out of their own understanding Reinforcement Learning • Reward based learning • Learning from +ve and –ve reinforcement • Machine learns how to act in certain environment • To Maximize rewards Kids will take actions on his own from past exp, and parent providing the feedback Decision Tree Random Forest Naïve Bayes Support Vector machineK-Means ClusteringTensorFlow MACHINE LEARNING ALGORITHM You Tube Recommendation videos System which play chess Linear Regression Logistic Regression
  • 6.
    MACHINE LEARNING –PROBLEM CLASSIFICATION Classification Algorithm Anomaly Detection Algorithm Regression Algorithm Clustering Algorithm Reinforcement Algorithm • Used to classify record • Limited number of answers • Analyze the pattern • Alerts in case of change in pattern • Used to calculate numeric value • Predicts outcome • Sepeartes the data into groups or clusters • Ease out the interpreation of data • Design how brains of humans respond • Learn from outcome , reward , punishment Email Spam for Gmail and Yahoo Prediction of house price QUESTIONS ML ALGORITHM ALGORITHM DESCRIPTION EXAMPLE Fraud Detection for Credit card Product Recommendation Google self drive car How much or how many ?3 Is this weird ?2 Is this A or B?1 How is this organized ?4 What should I do next?5
  • 7.
    Business Problem ML DEMO- PORTUGUESE BANK age job marital education default balance Housing loan contact day month duration campaign pdays previous poutcome Y 30 unemployed married primary no 1787 no no cellular 19 oct 79 1 -1 0 unknown No 33 services married secondary no 4789 yes yes cellular 11 may 220 1 339 4 failure no 30 management married tertiary no 1221 yes no telephone 25 jul 279 4 -1 0 unknown yes • There has been a revenue decline for the Portuguese bank and they would like to know what actions to take. • After investigation, they found out that the root cause is that their clients are not depositing as frequently as before. • Knowing that term deposits allow banks to hold onto a deposit for a specific amount of time, so banks can invest in higher gain financial products to make a profit. • In addition, banks also hold better chance to persuade term deposit clients into buying other products such as funds or insurance to further increase their revenues. • As a result, the Portuguese bank would like to identify existing clients that have higher chance to subscribe for a term deposit and focus marketing effort on such clients. To predict which clients are more likely to subscribe for term deposits Data Set Description Name Description and Values Personal Client Information Age Age at the contact date (Numeric ≥18) Marital Status Married, Single , Divorced, Widowed etc Sex Male or Female Job Unemployed , 'entrepreneur, technician Name Description and Values Bank Client Information Annual Balance in Euro currency (Numeric ) Default ( Credit) Yes , No , Unknown Housing Loan Yes , No , Unknown Pesonal Loan Yes , No , Unknown Name Description and Values Last contact information Contact Type contact communication type Date When the contact was made Duration Duration of the contact Name Description and Values Other Attributes campaign No of contacts performed during this campaign pdays No of contacts performed before this campaign poutcome Outcome of the previous compaign Output variable (desired target): y - has the client subscribed a term deposit? (binary: 'yes','no') Data Set Business Goal
  • 8.
    PYTHON DEMO STEPSFOR BANKING SCENARIO Import Python Library Packages Pandas , numpy, matplotlib Loading the Banking data and reading the file Drop the variables which are not required for prediction Spliting the data into training set and testing set Applied logistic regression model using training data set Checking the score or accuracy of the model Library File Data Set Splitting the Data Set Data Filter ML Model Accuracy DESCRIPTION PYTHON CODESTEPS
  • 9.
    REFERENCES Machine Learning book- Tom Mitchel The Elements of Statistical Learning book -Trevor Hastie Video Lecture - Andrew NG Video Lecture - Yaser Abu Mostafa Artificial Intellignece Reports from IDC , PWC , DB and McKinsey