Online | 6 months | 6 Terms

PROGRAM SNAPSHOT

PRE-TERM PREP TERM 04

Python and Stats for Machine Learning Intemediate
Data Science

TERM 05 [ ELECTIVE ]
TERM 01
Machine Learning Advanced - [E-1]
Data Analysis basics with Python
Tensorflow for Deep Learning -[E-2]

TERM 02
TERM 06
Data Visualization & EDA
Capstone Project - II &
Industry Immersion

TERM 03
Machine Learning Foundation Optional Terms
EDA with R
Data visualization with Tableau
PROJECT
Capstone Project - I
 Term 1:
DATA ANALYSIS BASICS WITH
PYTHON

Thought Experiment: Data Science from a layman's perspective

Module 1 : Brief intro to Data Science
Data Science How companies use Data Science
Fundamentals Overview of Data Science project lifecycle
Walkthrough of data types and data challenges

In class quiz for Python Basics

Module 2 : Common Python concepts and sample questions
Recap of Python
Variable, Inbuilt datatyes, functions, modules and Packages
for Data Science
File operations and error handling

In class quiz for Descriptive Statistics

Module 3 : Common charts used
Recap of Statistics In class quiz for Inferential Statistics
for Data Science Probability, Central Limit theorem, Normal Distribution & Hypothesis
testing

Introduction to Numpy Arrays

Module 4 : How to apply mathematical operations in Numpy
Mathematical Array manipulation using Numpy
operations Broadcast values across Arrays using Numpy
using Numpy

Types of Data Structures in Pandas

Module 5 : Clean data using Pandas
Data manipulation Manipulating data in Pandas
with Pandas How to deal with missing values
Hands-on: Implement Numpy arrays and Pandas Dataframes
 Project Expectation setting
Mini Project :
Project Timelines
Data manipulation
Github for Project Submission
Projects
Presentation guidelines

Term Projects

Indian Premier League 2008-2018 Car Sales Advertisment

Analyse key success factors for top cricket Analyse the car sale data in Ukraine.
team at IPL
The matches dataset contains 18 variables The dataset contains 10 variables and 9k+
and 600+ observations . The deliveries observations of the car sales data in Ukraine.
dataset contains 21 variables and 160k+
observations of the IPL 2018 season.

Olympic 1896- 2014 1000 movies data

Analyse which country have won the most Analyse the IMDB 1000 most popular movies
medals at Olympic games. and come up with interesting insights.
The dataset contains 9 variables and 31.2k The dataset contains 12 variables and 1000
observations of the summer olympic games observations of the top 1000 popular
(1896 - 2014) movies for past 10 years

Restaurant across America data

Analyse the availability and accessibility of

food across America

The dataset contains 10 variables and 10k

observation of the fast food restaurants in
America.
 Term 2:
DATA VISUALIZATION & EDA

Module 1 : Plotting basic statistical charts in Python

Data Visualization Data visualization with Matplotlib
in Python - 1 Case study: Analysis of Wine dataset through visualizations

Module 2 : Statistical data visualization with Seaborn

Data Visualization Interactive data visualization with Bokeh
in Python - 2 Case study: Analysis of Fifa data using Seaborn & Bokeh packages

Introduction to Exploratory Data Analysis (EDA) steps

Module 3 : Plots to explore relationship between two variables
Exploratory Histograms, Box plots to explore a single variable
Data Analysis - 1 Heat maps, Pair plots to explore correlations
Case study: Perform EDA to explore survival using Titanic dataset

Module 4 : Interaction with Industry experts

Industry immersion QnA
& project discussion

Module 5 :
Exploratory Case study: Analyse mental health of IT professionals
Data Analysis - 2
 Module 6:
Top 3 teams present their projects details to the entire cohort
Project presentation

Term Projects

Indian Premier League 2008-2018 Car Sales Advertisment

Analyse key success factors for top cricket Analyse the car sale data in Ukraine.
team at IPL
The matches dataset contains 18 variables The dataset contains 10 variables and 9k+
and 600+ observations . The deliveries observations of the car sales data in Ukraine.
dataset contains 21 variables and 160k+
observations of the IPL 2018 season.

Olympic 1896- 2014 1000 movies data

Analyse which country have won the most Analyse the IMDB 1000 most popular movies
medals at Olympic games. and come up with interesting insights.
The dataset contains 9 variables and 31.2k The dataset contains 12 variables and 1000
observations of the summer olympic games observations of the top 1000 popular
(1896 - 2014) movies for past 10 years

Restaurant across America data

Analyse the availability and accessibility of

food across America

The dataset contains 10 variables and 10k

observation of the fast food restaurants in
America.
 Term 3:
MACHINE LEARNING
FOUNDATION

What is Machine Learning ?

Module 1 : Use Cases of Machine Learning
Introduction to
Types of Machine Learning - Supervised to Unsupervised methods
Machine Learning (ML)
Machine Learning workflow

Introduction to Linear Regression

Use cases of Linear Regression
Module 2 : How to fit a Linear Regression model?
Linear Regression Evaluating and interpreting results from Linear Regression models
Case study: How linear regression helps determine demand

Introduction to Logistic Regression

Module 3 : Logistic Regression use cases
Logistic Regression Understand use of odds & Logit function to perform logistic regression
Case study:Predicting default cases in the Banking Industry

Introduction to Decision Trees & Random Forest

Understanding criterion(Entropy & Information Gain) used in Decision
Module 4 : Trees
Decision trees Using Ensemble methods in Decision Trees
& Random Forests Applications of Random Forest
Case study:Predict passengers survival in a Ship mishap

Introduction to evaluation metrics and model selection in Machine

Learning
Module 5 : Importance of Confusion matrix for predictions
Model evaluation Measures of model evaluation - Sensitivity, specificity, precision,
techniques recall & f-score
Use AUC-ROC curve to decide best model
Case study:Applying model evaluation techniques to prior case study
Term Projects

Predicting temperatures in Candy Classification

World War II
Classify if a candy is a chocolate or not based
Predict the maximum temperature when
on its features
minimum temperatue is known.
The dataset contains 31 varibales and 100k+ The dataset contains 13 variables and 86
observations of the weather conditions observations to predict the most or least
during World War II. popular Halloween candy.

Predicting Housing prices Predicting risk in Life insurance

Develop a predictive model that accurately
Predict final sales price of each house of
classifies risk, imapcting public perception of
residential homes in Iowa.
the industry.
The train dataset contains 81 variables and The train dataset contains 128 variables and
1461 observations. The test dataset contains 59.4k observations. The test dataset contains
80 variables and 1460 observations of the 127 variables and 19.8k observations of the
house. prices at any anonymous location. insurance applicants.

Credit Card Fraud Detection

Identify the fraudulent credit card transactions.

The dataset contains 31 variables and 285k

observations of transactions made by credit
cards in September 2013 by european
cardholders.
 Term 4:
MACHINE LEARNING INTEMEDIATE

Unsupervised Learning: Introduction to Curse of Dimensionality

What is dimensionality reduction?
Module 1 :
Technique used in PCA to reduce dimensions
Dimensionality
Applications of Principle component Analysis (PCA)
Reduction using PCA
Case study: Optimize model performance using PCA on high dimension
dataset

Introduction to KNN
Module 2 : Calculate neighbours using distance measures
KNN (K- Nearest Find optimal value of K in KNN method
neighbours) Advantage & disadvantages of KNN
Case Study:Classify malicious websites using close neighbour technique

Introduction to Naïve Bayes classification

Module 3 : Refresher on Probability theory
Naïve Bayes classifier
Applications of Naive Bayes Algorithm in Machine Learning
Case study : Classify Junk emails based on probability

Introduction to K-means clustering

Decide clusters by adjusting centroids
Module 4 :
Find optimal 'k value' in kmeans
K-means
Understand applications of clustering in Machine Learning
clustering technique
Case study : Segment flower species in Iris flower data

Introduction to SVM
Module 5 : Figure decision boundaries using support vectors
Support vector Identify hyperplane in SVM
machines (SVM) Applications of SVM in Machine Learning
Case Study : Predicting wine quality without tasting the wine
 Introduction to Time Series analysis
Stationary vs non stationary data
Module 6 : Components of time series data
Time series
Interpreting autocorrelation & partial autocorrelation functions
forecasting
Stationarize data and implement ARIMA model
Case Study: Forecast demand for Air travel

Term Projects

Simplified Human Activity

SMS Spam collection data
Analyse the recordings of subjects performing
Classify collection of spam messages tagged as activities while carrying inertial sensors.
spam or legitimate. The test dataset contains 562 variables and 1542
observations. the train dataset contains 563
The dataset contains 5 variables and 5572 variables and 3609 observations of subjects
observations collected for SMS spam performing activities while carrying inertial
research. sensors.

Gender recognition by voice Store Item Demand Forecasting

Identify a voice as male or female(SVM) Predict 3 months of item sales at different store.

The dataset contains 21 variables and 3k+ The test dataset contains 4 variables and
observations to identify a voice as male or 45000 observations. the train dataset
female using acoustic properties of voice and contains 4 variables and 90k observations of
speech. a 5 year of store-item-sales data.

Safe driver prediction

Predict the probability that an auto insurance
policy holder files a claim.
The test dataset contains 58 variables and
80k+ observations. the train dataset contains
59 variables and 60k+ observations of
driving data.
 Term 5 [ Elective ]:
MACHINE LEARNING ADVANCED

Applications of Apriori algorithm

Module 1 :
Understand Association rule
Introduction to
Developing product recommendations using association rules
Apriori Algorithm
Case study : Analyse online data using association rules

Introduction to Recommender systems

Types of Recommender systems - collaborative, content based &
Module 2 :
Hybrid
Recommender
Types of similarity matrix (cosine , Jaccard, Pearson correlation)
Systems
Case Study:Build Recommender systems on Movie data

Recap of dimensionality reduction concepts

Module 3 : Types of dimensionality reduction
Linear Discriminant
Dimensionality reduction using LDA
Analysis (LDA)
Case Study : Apply LDA to determine Liquor Quality

Introduction to Anomaly detection

Module 4 : How Anomaly detection works?
Anomaly Detection
Types of Anomaly detection: Density based, Clustering etc.
Case Study:Detect anomalies on health data

Introduction to Ensemble Learning

Module 5 : What are Bagging and Boosting techniques?
Ensemble learning What is Bias variance trade off?
Case study : Predict wage (annual income) classes from adult census
data
 Introduction to stacking
Module 6 : Use Cases of stacking
Stacking How stacking improves machine learning models?
Case Study:Predict survivors in Titanic case

Introduction to optimization in ML
Applications of optimization methods
Module 7 :
Optimization Optimization techniques: Linear Programming using Excel solver
How Stochastic Gradient Descent(SGD) Works?
Case study: Apply SGD on Regression data (sklearn dataset)

Introduction to Neural networks

Module 8 : What are Perceptrons & Types of Perceptrons?
Neural Networks Workflow of a Neural network & analogy with biological neurons
Case Study : Apply computer vision for digit recognition on MNIST data

Term Projects

Apriori Algorithm(Market MovieLens Dataset

Basket Analysis) Predict the name of movies and based upon the
Market Basket Analysis of e-commerce data of reviews of the other critics having similar taste.
transaction of 2010 and 2011 The combined dataset consists of 4 different dataset.
The links dataset have 3 variables and 9k+ observation,
The dataset contains 8 variables and 542k observa- the movies dataset have 3 variables and 9k observation.
tions of all the transaction of 2011 and 2011 for a The ratings dataset have 4 variables and 100k observa-
UK based and registered non - store online retail. tion. The tags dataset have 4 variables and 1297
observations.
 Pokemon Dataset(LDA) Property Inspection prediction
Build a pokemon dream team of 6 pokemon that
Predict a transformed count of hazards or
inflicts the most damage while remaining
pre-existing damages using dataset of property
impervious to any other team of 6 pokemon.
information.
The dataset contains 41 variables and 801 Predict a transformed count of hazards or pre-ex-
observation of data on pokemon from all 7 isting damages using dataset of property informa-
generations. tion.

Letter recognition

Predict the letter category based on its attributes.

The test dataset contains 17 variables and
4000 observations. the train dataset contains
18 variables and 16k observation of 26 capital
letters of english alphabet based on their
different attributes.
 Term 5 [ Elective ]:
TENSORFLOW FOR DEEP LEARNING

Introduction to Artificial Intelligence

Module 1 : Breakthroughs in the field of AI
Artificial Intelligence
Overview of advanced Machine Learning algorithms
Weights and Bias estimation using gradient descent optimization

Installing Tensorflow in Python

Introduction to data flow graphs in Tensorflow
Module 2 :
Getting started Functions, operations and execution pipeline in Tensorflow
with Tensorflow Regression technique in Tensorflow
Case Study: Predict Boston Housing Prices using Tensorflow

Module 3 : Classification in Tensorflow

Tensorflow Introduction to Tensorboard visualization
programming
Activation functions in Tensorflow
in Python

An overview on Deep Neural Networks

Module 4 : Real world applications of Deep Neural Networks
Introduction to Neural Networks using Tensorflow
Deep Learning Optimization techniques employed in Neural Networks
Case Study: Classify handwritten digits (MNIST) using logistic regression

Hyperparameters in deep neural networks

Module 5 : Filters in Convolutional Neural Networks
Optimization of Deep
Max pooling and padding
Neural Networks
Dropout and Regularization in Deep learning
 Introduction to Convolutional Neural Networks(CNN's)
Module 6 : Evaluate, Improve and tune Convolutional Neural Networks
Introduction to Object Classification, localization and segmentation
Convolutional Neural Reusing models with Transfer learning
Networks Case Study: Classify handwritten digits (MNIST) using Deep Neural
networks with RELU on Keras

Module 7 : Introduction to NLP

Introduction to Introdution to Word embeddings
Natural Language Simple Word Vector representations: word2vec, GloVe
Processing Implementation of word2vec model in Keras

Intoduction to Recurrent neural networks based language models

Module 8 : Introduction to Gated Recurrent units
Recurrent Neural LSTMs for machine translation
networks Case Study: Perform Sentiment Analysis using word embedding
Seq2Seq LSTM Model translation

Introduction to Sequence to sequence learning

Module 9 : Convolutional Neural networks for Sentence classification
Recursive neural
Train Recursive Neural Networks for Sentiment analysis
networks for
Introduction to Dynamic memory networks
Sentiment Analysis
Case Study: Explore Pride and Prejudice book to perform Char-RNN

Module 10 :
Project: Bulid a Chatbot using Slack Class
Revision of concepts
Integrate Chatbot with Bot server
and Term Project
Term Projects

Build Chatbot using slack Class

Develop a chatbot to simulate a human as

a conversational partner

This chatbot allows us translate user

submitted conversations from English to
Hindi
 Term 6:
CAPSTONE PROJECT - I

James Telco Bond

In this capstone project, students will be provided with data collected by a major
Telecom operator on the demographic behaviour of users using different
handsets.

Students are required to do the initial bit of data cleansing, pre-processing and
then upload this data to SQL server via a web hosting platform that will be
provided to them.

This data from SQL server will be used to create a dashboard for the
company using D3.js scripts. D3,js scripts will be provided to students upfront.
These dashboards are reflective of how interactive visualizations can help
companies make strategies such as what demographies to cater to, how men and
women customers behave differently, which geographies are popular and ones
that need more investment from the company in terms of finance and marketing?
 Term 6:
CAPSTONE PROJECT - II

Demand Planners

This capstone project will focus more on applying machine learning concepts
rather than data gathering and storing aspects. Students will be provided with
data collected by a major Taxi Aggregator of taxi bookings done in a leading city.
As budding data science consultants, students are required to do exploratory data
analysis & present an initial report.

After that the students are required to create an UI that displays the
observations regarding taxi usage across the city from the analysis and the
website should also have a provision for the company to forecast demand for
taxis at a specific time in the day.

The taxi bookings data provided will be in csv format and dashboards for the
company need to be created using D3.js scripts. The D3.js scripts will be provided
to the students beforehand.
 OPTIONAL TERMS
EDA with R

Thought Experiment: Data science @ Google

Module 1 : Introduction to Data Science
Data Science Real world use-cases of Data Science
Fundamentals Walkthrough of data types
Data Science project lifecycle

Installing R and R Studio

Basic Commands in R
Module 2 :
Introduction to Installing packages
programming in R Setting working directory
Exercises: Basic exercises in R Programming

Data structures
Module 3 : Basic Data management
Playing around with Loops and Functions
Data objects in R Saving output
Exercises: Loops and functions in R

Introduction to Statistics
Module 4 : Descriptive Statistics
Descriptive Measures of central tendency
statistics - 1 Measures of Dispersion and shape
Case Study: Investigation of Crime statistics in Beaufort

Introduction to Probability
Module 5 : Probability Distributions used in Data Science
Descriptive
Quantiles, percentiles, and standard score
statistics - 2
Case Study : Analyse student's performance at school
 Introduction to Inferential Statistics
Module 6 : Population and Samples
Inferential
Central Limit Theorem
Statistics - 1
Case Study: Sampling data for Business analysis

Introduction to Hypothesis Testing

Module 7 : Confidence Intervals
Inferential Tests of significance: p-value
Statistics - 2 Case Study: Apply Inferential statistics & Central limit theorem using
Python

Loading data from R libraries

Module 8 : Importing data from Excel and CSV files
Intermediate R: Connecting SQL databases
Importing data Webscraping using R
Case study: Webscraping websites using scrapy package

Identifying NULL values in datasets

Module 9 : Introduction to data imputation methods
Intermediate R : Data Creating new variables and recoding variables
Manipulation using Type conversions
Tidyverse Case Study: Using Tidyverse in Data Manipulation

Managing Date values

Module 10 : Numerical and Character functions
Intermediate R: Aggregating & Restructuring data
Restructuring Data Sorting, merging datasets
Exercises : Subsetting datasets for use in Predictive analytics
 Module 11 : Introduction to basic graphs: Barplots, Scatterplots & line graphs
Intermediate R : Using Boxplots in univariate analysis
Exploratory data Applications of Histograms
analytics using ggplot2 Using ggplot2 for advanced visualizations

Term Projects

Kickstarter funding patterns How bad is the Air Quality in

metropolitans?
Derive insights on successful and failed projects Analyse worsening airquality in metropolitan
on Kickstarter platform cities
The dataset contains 6 variables and over
The dataset contains 15 variables and
100 observations
around 400,000 observations

Marketing strategies in What are the Characters in

Retail banking superhero comics
Derive insights on how sucessful are the Identify the good, bad and the ugly nature of
direct marketing campaigns of a Portuguese characters in Marvel comics
Bank
The dataset contains 11 variables and over
The dataset contains 17 variables and over
20000 observations
45000 observations

What caused International crisis

Identify all factors which caused major
international crisis events in the last 100
years
The dataset contains 96 Variables and over
1000 observations
 OPTIONAL TERMS
DATA VISUALIZATION WITH TABLEAU

Introduction to data visualization

Module 1 : Understanding Tableau ecosystem in industry
Introduction to Loading data files in Tableau
Visual Analytics Creating first visualizations
Case Study: Sales performance Analysis

Introduction to graphs - bar graph and line graph

Module 2 : Working with continuous measures & discrete variables
Data Visualization Heat maps and Geographical data visualizations
using Tableau Creating map Views
Case Study: Analyse Natural calamity trend and effect

Introduction to SQL joins

Module 3 : Performing data blending in Tableau
Data joining & Creating dual axis charts in Tableau
blending in Tableau Introduction to descriptive statistics and Visual analytics
Case Study : Analyse trends in Retail businesses

Introduction to R programming tool & R studio

Module 4 : Installing R and R studio
Predicitve Analytics Applications of linear regression in prediction
using Tableau and R Data crunching: Creating groups, sets & parameters
Case Study: Forecast revenues

Introduction to principles of dashboard design

Module 5 : Custom geocoding in Tableau
Interactive Developing dashboard products using Tableau
Dashboard Design Introduction to writing storyline in Tableau
Case Study: Customer segmentation dashboard
 Introduction to calculations: Date calculations
Using LOD calculations: INCLUDE, EXCLUDE & FIXED functions
Module 6 :
Working with Table calculations
Advanced Calculations
Exporting data from Tableau
using Tableau
Case Study: Analyse sales across geographies, products & customers

Introduction to customer churn analysis

Module 7 :
Estimating customer life time value
Applications of
advanced Calculations Applications of context filtering
using Tableau Applications of logical functions in Tableau
Case Study: Analyse retail sales data to predict customer behaviour

Revision of key concepts: data blending, writing calculations, LOD

Module 8 : calcuations etc.
Revision of Review of Tableau project portfolio
concepts and Communicating data insights using reporting tools
Project discussion Tableau Interview prep
Discussing EDA objectives of final project

Term Projects

Hubway data visualization

challenge
Produce visualizations that reveal interesting
user patterns about how people in Boston
gets around on Hubway
The dataset contains 1 million observations
on bike usage by residents of Boston
Need to know

Program Start August 2018

Duration 06 months (Incl. Capstone Projects)

Prerequisite Background in Programming (Not Mandatory)

Laptop with 4 GB RAM

Program Fee INR 2 Lakh + GST

Scholarships 50 scholarships (each with 70% tuition waiver) for

professionals passionate about making a career in
Data Science & furthering INSAID’s mission of
putting India on the global AI map.

Talk to our Admissions team today or attend the next Data

Science MasterClass to know more

For Further details, write to us [email protected]