Data Science in Practice - An Online Bootcamp

This sub-module will enable participants to differentiate real data science from the fuzz around it and develop a solution-oriented mindset. It is scheduled in the beginning of the workshop. • The four building blocks of Data Science • Common mistakes and best practices for data scientists • Relation of Data Science with AI, Machine Learning, Deep Learning, and other fields • Technology stack and choosing the best technology • An overview of industrial applications and their requirements

Scheduled in the middle of the workshop, this module will introduce participants with practical requirements of business use-cases and best practices during CRISP-DM or garage approach. • Identifying and prioritizing data science use-cases within an organization • Translating business problems into ML problems • Analytics-roadmap for organizations • Mapping of Design Thinking to Data Science • Introduction and requirements for ML project P2, that will follow this session

At the end of the workshop when the participants have gone through the end-to-end data science journey, this module will help participants understand which role fit best to them, the best practices, and how to enter and excel in that role. • Data Science value perspective • Data Science teams and organizations • Data Scince roles (e.g. Data Engineer, Analyst, Data Scientist, ML Engineer, etc) • Methods such as CRISP-DM, garage, scrum • Communication with non-data scientists • The non-technical skills needed for excellence in Data Science

Python is an easy to understand scripting language – yet its compact programming style and vast amount of libraries makes it a challenge for learners to focus on what matters the most. With a carefully designed contents trajectory, this hands-on module will provide participants a good basis in Python for the rest of the workshop – which then includes most relevant libraries and methods. • Why Python is the most popular language for Data Scientists • Introduction to Python as a language • Python native data structures including Lists, Set, Dictionary, Tuple • Numpy and Scipy • Control Structures • Functions and Classes • Hands-on Exercises in Python

EDA is what consume most time of a Data Scientist (50% to 80%). This module will equip participants with best techniques to do efficient and effective EDA. • Introduction to EDA, common mistakes, best practices • Introduction to Pandas library • DataFrame and Series data structures in pandas • Reading data from different sources (csv, web, excel, json, SQL, txt, etc) • DataFrame operations e.g. filtering, filling, merging, conditioning, aggregation • Summarization, outlier detection, and bird-eye-view reporting • Map-reduce for efficient operations on DataFrames • EDA hands-on excercises

Data Visualization is not only helpful to communicate results and findings with others but is equally important for data scientist itself in order to understanding the data. This module will enable participants to develop quick and pretty visualizations using Python libraries. • Good and bad types of visualizations • Practical working with Matplotlib – making any visualization • Practical working with Seaborn – making interactive and pretty visualizations • Practical working with Plotly – making and deploying interactive and pretty visualizations

This module will allow participants test and improve their skills developed so far i.e. in EDA with pandas, and visualization libraries in Python. A business use case will be discussed as a reference case for exercises. • Use-case and problem statements • Data loading and merging • Data analysis • Data cleaning • Data exploration • Data visualization

This module will provide theoretical understanding of machine learning algorithms, their working, their advantages and limitations, and hence demystifying it for participants – eventually hoping they will be able to decide their own design of ML if needed. • Types of Machine Learning algorithms and application scenarios • Classification algorithms (e.g. Naïve Bayes, Decision Trees, KNN, ANN, Support Vector Machines) • Regression algorithms (e.g. cousins of classification algorithms, Linear/Ridge/Lasso Regression and all other cousins of classification algorithms) • Ensemble methods (e.g. Random Forests, Gradient Boosted Trees) • Outlier detection (e.g. One-Class SVM, auto-encoders) • Clustering algorithms (e.g. K-Means, DBSCAN, Hierarchical clustering) • Feature Selection and Dimensionality reduction (e.g. PCA, LDA, RFE and other techniques)

By using Scikit-learn as the main library, this tool will enable participants to apply machine learning process through model selection, model building, parameter optimization and evaluation. • Introduction of Scikit-learn • Exploration of built-in datasets • Building an ML model using Scikit-learn • Splitting data into training, validation, and testing • Cross-validation techniques • Hyperparameter search using GridSearchCV • ML Excercises and Assignment

This is the first of two projects in Machine Learning. The objective of this projects is to enable participants apply their so far gained knowledge of end-to-end data science process (EDA+ML) on a real-world scenario. • Quick introduction of the use-case • Loading data from different sources • EDA and Data cleaning • Building machine learning models • Validation and testing of models • Debugging machine learning model w.r.t overfitting and underfitting

Primary objective of this module is to enable participants to use Pipelines as end-to-end machine learning construct. The module will also cover more details on feature selection, and dimensionality reduction with exercises. • Curse of Dimensionality • Feature selection methods (Univariate and multivariate methods) • Dimensionality reduction techniques exercises (PCA, LDA, RFE, etc) • Concept of transformations and operations in Python • Introduction to Machine Learning Pipelines in Scikit-learn • Building 2-steps, 3-steps, k-steps ML Pipelines (e.g. Feature Selection + Feature Engineering + Classification) • Pipelines and GridSearchCV • Exercise of all of above • The untold truth of Machine Learning

This is second ML project with and objective to exercise a full CRISP-DM cycle i.e. domain-understanding + data understanding + data analysis + model building + model optimization + model deployment (as a pipeline). • Domain understanding from a problem statement • Exploratory Data Analysis • Feature Engineering and Feature Selection • Building machine learning models • Validation and testing of models • Doing it all within a Machine Learning Pipeline • Applying CRISP-DM

This is second ML project with and objective to exercise a full CRISP-DM cycle i.e. domain-understanding + data understanding + data analysis + model building + model optimization + model deployment (as a pipeline). • Building ML Apps with REST APIs using Flask • Introduction to Apache Spark • Machine Learning on Spark – A Hands-on session • Introduction to Cloud-based Artificial Intelligence • Architecture of large-scale AI Applications

This module will enable participants understand and experiments on the concepts of decentralized applications. • Understanding Blockchain • Distributed Ledgers • Cryptocurrencies • Blockchain Potential: Use cases • Building a Blockchain application with Python (hands-on)

Deep Learning is probably the most talked-about area in Data Science. This module will equip participants with the understanding and practical experience of building a deep neural network using Keras/Tensorflow. • Types and applications of Neural Networks • Multi-layer Backpropagation Networks • Activation function (Sigmoid, Tanh, Relu, etc) • Introduction to Convolutional Neural Network • Introduction to Keras • Development of Image classification using Deep Neural Network • Development of Image classification using Convolutional Neural Network