Practical Data Scientist Online Program

The goal of this module is to introduce the basic framework of data science called Cross Industry Standard Process for Data Mining (CRISP-DM). Learners will understand the philosophy behind the data science framework. In addition to that, this module will also delve into the critical issue of converting business problem statements into data problems.

This module introduces some of the modern tools and techniques of software development such as version control using Git. It would also be helpful to have an understanding of the Agile processes in a Data Science project.

• Local
  • Identify the programming language (Python, R, Julia etc)
  • Evaluate the IDEs (Jupyter, PyCharm, RStudio, VSCode etc)
  • Version Control using Git (optional)
  • Setting up codebase in Bitbucket (or Github)
  • Introduction to REST APIs
• Cloud

  In this module gain immense knowledge of Cloud Computing and disadvantages of on-premise infrastructure, Deploying a Machine Learning Model end to end on the cloud using Amazon Web Services like Cloud Formation, Lambda, S3 and Machine Learning Services used in various Projects.

  • Create an AWS Account

    One need to understand cloud computing and essential concepts of Cloud Computing - Cloud Deployment Models, Cloud Service Models. Get an overview of AWS Global Infrastructure, Regions and Availability Zones

  • Setup your IAM Role

    Understanding the security features of AWS through IAM service through Users, Groups Roles and Policies

  • Create an S3 bucket (storage)

    Learn how to use storage services in AWS through S3, Creation of a Bucket, Advantages of S3, Properties of S3, Storage Classes. Connecting S3 to another AWS service, Building a Data Lake using S3.

  • Create a SageMaker instance

    Gain Broad idea about Machine Learning on Cloud, Sage Maker as a Service, Various sub services under Sage Maker, Creation of Sage Maker Notebook Instance, working with Jupyter Notebook instance, Overview of Sage Maker Studio, Building a Model in Jupyter, Deployment of the Final Model

  • Amazon Kinesis Data Stream and Firehose

    To Learn collect process and stream large streaming data, Firehose which helps to deliver the data into S3

  • Cloud Formation

    Deploy Quickly your required AWS services using Cloud Formation

  • Amazon API Gateway

    Understanding API and applicability of the Amazon API Gateway Service to deal with Creating, Maintaining, Monitoring and Securing APIs.

SQL, NoSQL, NewSQL, Cloud Storage This module introduces the databases that typically exist in the business environment. Traditional (Structured data) databases like Oracle, MySQL, SQL Server, DB2, etc, and the query language (SQL). We also get used to the NoSQL databases (Unstructured data) like HBase, MongodB, Cassandra, CouchDB, etc. Understand the architecture and design of the new age databases capable of handling new age data requirements based on consistency, availability and partition tolerance.

• Data Models/Formats
  • Structured Data
  • Semi Structured Data
  • Unstructured Data
• Data File Formats
  • Text/CSV
  • JSON
  • Sequence Files
  • AVRO Files
  • Parquet
  • RC Files
  • ORC file format
• Types of Databases
  • SQL (MySQL / Amazon RDS)
  • NoSQL & NewSQL
    • key-value store (Redis)
    • Document store (MongoDB)
    • Column-oriented (HBase)
    • Graph (Neo4j)
  • Cloud Storage (DynamoDB / S3)

This module will get the users up to speed on the programming requirements of being a Data Scientist. Python is emerging as the language of choice for Data Scientists but interested candidates can also choose to opt for R language. In the Python programming track, Object oriented programming concepts are introduced as well.

• Course Introduction and Python installation/setup environment
• Basic Python Concepts
  • Printing
  • Strings
  • Data types
  • Numeric Operators
  • Slicing and Dicing
  • String Operators
• Flow Control
  • If, elif and else operators
  • Conditional Operators
  • While loops
  • For loops
  • Break, nested loops
• Tuples, Ranges and Lists
  • SQL (MySQL / Amazon RDS)
  • NoSQL & NewSQL
    • key-value store (Redis)
    • Document store (MongoDB)
    • Column-oriented (HBase)
    • Graph (Neo4j)
  • Dictionaries and Sets
    • Operations on Dictionaries
    • Sets Operations
  • Input and Output in Python
    • Reading and Writing text files
    • Pickling (Serialization) files
    • Understanding Shelve (Data storage persistence)
  • Using Databases in Python
    • Introduction to Databases and Terminology
    • Installation of Sqlite3
    • Querying data using SQLite
    • Joins, Complex joins
    • Exception handling
    • Working with NoSQL and NewSQL databases
  • Object Oriented Programming using Python
    • OOP concepts - classes
    • Instances, Constructors and more
    • Methods
    • Inheritance
    • Polymorphism
    • Composition
    • Aggregation
    • Decorators

This module begins to set up the groundwork for the core skills of being a Data Scientist by introducing the learning to basic statistics. We will discuss probability distributions, descriptive and inferential statistics.

• Data types
  • Continuous, Discrete, Categorical, Count
  • ominal, Ordinal, Interval, Ratio
• Introduction to Probability
  • Random variable
  • Probability and Probability Distribution Function
  • Balanced vs Imbalanced datasets
  • Sampling techniques for handling imbalanced data
  • Sampling Funnel - population, sampling frame, simple random sample
• Introduction to statistical concepts
  • Expected value of a probability distribution
  • 1st moment - measure of central tendency (mean, median, mode)
  • 2nd moment - measure of dispersion (Variation, Standard Deviation, Range)
  • 3rd moment - Skewness
  • 4th moment - Kurtosis
• Graphical tools for statistical analysis
  • Bar plot
  • Histogram
  • Box Plot
  • Scatter plot
• Normal Distribution
  • Introduction
  • Standard normal distribution or Z distribution
  • Z scores and Z table
  • QQ plot and QQ table
• Advanced statistical techniques
  • Sampling variation
  • Central limit theorem
  • Sample size calculator
  • T-distribution and student’s T-distribution
  • Confidence interval

After gaining a basic introduction to statistics, this module will introduce Hypothesis testing and Analysis of Variance (ANOVA) and other useful statistical concepts.

• Parametric vs Non-Parametric tests
• Formulating a hypothesis
• Choosing Null and Alternative Hypotheses
• Type I and Type II errors
• Comparison of sample proportions using hypothesis testing
• 2 sample t-test
• 1 sample t-test
• 1 sample z-test
• ANOVA
• 2 proportion test
• Chi-square test
• Non-parametric test

• Simple Linear regression
  • Correlation analysis
  • Correlation coefficient
  • Ordinary least squares (OLS) regression
  • Split data into train, test and validation sets
  • Overfitting (Variance) vs Underfitting (Bias) trade-off ratio
  • Generalization error and regularization techniques
  • Heteroscedasticity
• Multiple regression
  • LINE assumption
  • Collinearity (Variation Inflation Factor, VIF)
  • Normality
  • Model quality metrics
  • Deletion Diagnostics
  • Logistic regression
    • Types of logistic regression
    • Assumptions and Steps of logistic regression
    • Multiple Logistic regression
      • Confusion matrix
      • Receiver Operator Characteristic (ROC) Curve
      • Lift charts and gain charts
    • Discrete probability distribution
      • Binomial distribution
      • Negative binomial distribution
      • Poisson regression
    • Advanced Regression
      • Poisson regression
      • Poisson regression with offset
      • Negative binomial regression
      • Zero inflated models
    • Multinomial regression
      • Logit and log likelihood
      • Category baselining
      • Modeling nominal categorical data
      • Lasso and Ridge regression

This module is one of the most interesting, laborious and creative parts of the total model development process. It deals with understanding the data, visualizing it to find correlations and begins the process of getting the data ready for use by various machine learning algorithms.

• Importance of visualization
  • Principles of visualization
  • Tufte’s graphical integrity rule
  • Tufte’s principles of analytical design
• Basic visualization techniques
  • Scatter plot
  • Area plots
  • Histograms
  • Bar charts
• Specialized visualization techniques
  • Pie charts
  • Box plots
  • Bubble plots
• Advanced visualization techniques
  • Waffle charts
  • Word clouds
  • Heatmaps
• Visualizing geospatial data
  • Introduction to Folium
  • Maps and markers
  • Choropleths

This module is an important part of the data science lifecycle because it determines how features can be extracted from the dataset to maximize the output from machine learning algorithms.

• Data cleansing
  • Handling missing and null values
  • Imputation techniques
  • Handling duplicates
  • Outlier analysis
• Feature selection
  • Correlation analysis
  • Using Lasso and Ridge regression
• Feature transformation
  • Log transformation
  • Scaling
  • Binning
  • Categorization
  • Handling date time fields
• Dummy variables
• Encoding
  • One hot encoding
  • Label encoding

This module introduces the popular machine learning algorithms that are used by data scientists for model development. Since this is a vast subject, we focus on just using a few examples of each paradigm of machine learning (supervised, unsupervised etc)

• Unsupervised
  • Clustering (k-Means, Hierarchical Clustering)
  • Segmentation
  • Principal Component Analysis
• Supervised
  • Decision Tree
  • Bagging and Boosting
  • Random Forest Model
  • Support Vector Machines
  • kNN
  • Gradient Boosting
  • eXtreme Gradient Boosting (XGBOOST)
  • Ensemble Techniques
•  

This module is a course in and of itself, but for the purposes of this course, we will review at a high level some of the most popular deep learning frameworks using Tensorflow, Keras and PyTorch.

• Multilayer Perceptron
• Backpropagation and Feedforward Architectures
• ANN parameters
• Convolutional Neural Networks (CNNs)
• Autoencoders
• Recurrent Neural Networks (RNNs)
• Long Short Term Networks (LSTMs)
• Regularization Techniques
• Generative Adversarial Networks (GANs)

• Understanding how to retrieve data from various data sources.
• Learn to extract the structured or unstructured data from various data sources to perform batch and real time processing.
• Learn the different best practices for processing data extracted from cloud platforms and on premise data sources. Understand the pros and cons of ingestion tools

• Sqoop: SQL to Hadoop (vice versa)
• Flume: Ingestion of log data
• Storm: Continuous stream data converted into batch data
• Kafka Cluster: Real time Data Ingestion (Streaming Data)
  • Producer
  • Consumer
  • Streams
  • Connector
• Spark Streaming - Near real time data processing from IoT devices
  • Spark Streaming Context
  • Spark window (Time Interval for collecting batch of Data)

Finally, we develop and evaluate a model. This will usually be an iterative process, where multiple models are developed and tested for effectiveness. Model evaluation techniques are introduced and the best practices are outlined.

This module describes the CI/CD pipeline to deploy models in the cloud environment using Jenkins (AWS/GCP).

• Creates a fully managed build service that compiles source code
• Checks for any new changes on GitHub every two minutes
• Zips the files and sends them to a predefined Amazon S3 bucket
• IAM S3 bucket policy - Allows the Jenkin server access to the S3 bucket
• S3 policy enables the HTTP request plugin of Jenkins server to access the S3 bucket

AWS:

Brief introduction to

• S3
• Lamda
• Batch
• EC2
• SageMaker
• EMR - Distributed Computing
• EKS
• ECR
• IAM
• CloudFormation

Using all the above services, build an end to end machine learning pipeline that runs in a fully managed production environment.

Finally, this module wraps up the course by describing the best practices on how to effectively monitor the models in production and when to retrain them.

• Amazon SageMaker model monitor enables us to capture the input, output and metadata for the invocations of the models that we deploy.
• We can use it to analyze the data and monitor its quality. With S3 for data storage.
• Amazon SageMaker makes it easy to efficiently extract and analyze the data.
• It detects when the performance of a model running in production begins to deviate from the original trained model.
• Amazon SageMaker Model Monitor alerts developers when drift is detected and helps them visually identify the root cause.