Certification Program in Data Science

Exploratory Data Analysis is usually where the data scientists tend to spend most of their time in the project. This often involves understanding the dataset, summarizing and describing the data at hand.

This is the phase where the statistical techniques are applied to draw inferences from the data. Typically, there are a lot of statistical techniques that are leveraged during this module and applied to the data.

This module introduces the various visual plots such as the bar plot, histogram, scatter plot, box and whiskers plot, etc., and how they can be leveraged to identify patterns and correlations in the data.

This module introduces the basic concepts of probability distributions and how that knowledge is extensively applied in data science projects. The various types of distributions like Gaussian (Normal), Bernoulli, Poisson, Binomial, Multinomial, etc.

This module introduces the concepts of Hypothesis testing and defines in detail about Null and Alternative Hypothesis and the scenarios where each of them could be proved. It discusses in detail the concepts like parametric tests like 1 sample t-test, 2 sample t-test, ANOVA and nonparametric tests like Chi-Square tests.

This module marks the beginning of predictive analytics and lays the foundation for the rest of the course modules of Machine Learning. The module introduces one of the oldest and simplest supervised learning techniques - the linear regression (ordinary least squares).

As can be easily understood, this is an extension of the OLS Linear Regression, by expanding the input space into multiple linear outputs. We discuss this using one of the most commonly used datasets and learn how it can be implemented in Python and R programming languages.

This module discusses the other advanced and useful types of regression models - Lasso and Ridge regression. We discuss the pros and cons of each of these models with practical examples using Python and R.

Also called the Logit model, this algorithm predicts the probability of the event falling into a certain category - Pass/Fail, Fraud/Not Fraud, Yes/No etc. We also get introduced to the concept of maximum likelihood estimation techniques which form the regression coefficients.

This is an extension of the Logistic Regression - the only difference being the output class could have more than 2 classes.

Count data is a data type in which the observations can only consist of non-negative numbers (0,1,2, 3 …). This module introduces regression models for such data such as Poisson Regression and Negative Binomial Regression and their application in survival analytics.

This module provides a general overview of unsupervised learning before launching into one of the most famous of its techniques - Clustering. Multiple algorithms such as connectivity based (Hierarchical), centroid based (k-Means), Distribution based, etc are discussed and implemented along with real- life scenarios.

This module discusses another great technique called the Principal Component Analysis. It is one of the dimensionality reduction algorithms which answers the question - how do we reduce the feature space without losing a lot of information.

This is a rules based machine learning algorithm where the algorithm learns interesting patterns in data and attempts to answer the question - what goes with what. Different algorithms such as Apriori, FP-growth are discussed and implemented.

This module builds on the concepts of association rule learning and implements a recommendation engine. A great example could be the recommendation engine as seen in Amazon or any other e-commerce platform that suggests items based on your current or previous selections.

This module builds on the fundamentals of Graph Theory and delves into the architecture and implementation of any network with emphasis on Social Networks. The module extensively covers the modelling and visualization of networks and some practical implementations.

This module introduces machine learning algorithms by discussing one of the most popular algorithms - the K-nearest neighbor algorithm which could be used both for classification and regression.

This module introduces the decision tree algorithm which is a popular non-linear tree based algorithm. Furthermore, the discussion then introduces the concept of Bagging which is a type of ensemble technique and Random Forest algorithm which is a type of bagging algorithm.

This module builds on the Bagging algorithm introduced earlier and also adds another ensemble technique called Boosting. The theory behind both the ensemble techniques is discussed in detail.

This module introduces Gradient Descent and explains how it is combined with boosting to achieve gradient boosting. Then we discuss the advanced implementations of gradient descent such as AdaBoost and Extreme Gradient Boosting.

All the prior modules have been dealing with structured data (data that can be stored as rows and columns and that follow relational requirements). This module introduces the unstructured data in the form of text data and provides some tools and techniques to analyze it.

This module discusses another important algorithm - Naive Bayes model which is based on the Bayesian Probability Technique. This algorithm has been successfully deployed to detect spam with great accuracy.

This module discusses the perceptron, which is the fundamental concept of an artificial neural network. The basic version of an ANN called the multi layer perceptron (MLP) is also discussed.

This module discusses the various building blocks of neural networks - perceptron, backpropagation, activation functions, dropout, dense vs sparse layers ,etc.

The difference between ANN and Deep Learning is that the network in deep neural networks consists of multiple hidden layers vs just a single layer in the ANN. The architecture is explored and some guidelines are laid out.

This module discusses another Black Box technique - Support Vector Machines.

This module builds on the concepts from Module 12 and discusses how survival analysis is performed in practical and real-life scenarios.

This module introduces the concept of a time series. All the previous modules (except text data) were longitudinal or cross-sectional datasets. There is no temporal component in the data (or it is ignored for the purposes of algorithm). This module introduces techniques to deal with time-series data such as AR, ARMA and ARIMA models.

This module builds on the concepts previously discussed and delves into some of the data-driven algorithms available to address the time-series problems such as MA, EMA and Econometric Models.