Data Science & AI

Understand various data sources and why organizations are gearing up to store the data like never before. Learn on what are the various applications of data science in various industries ranging from FSI to LSHC to Retail and many more. Also one will appreciate the job opportunities in the space of data science, data modeling, and data analysis. Finally understand the golden rule on how to become a successful data scientist, data modeler, data analyst, etc.

Learn about the Project Management Methodology, CRISP-DM, for handling Data Science projects and various concepts used in defining business problems and then performing data collection in line with business problems. Understand the importance of documenting the business objectives and business constraints so that the entire project is performed to solve business problems. Project charter overview will help participants understand the real-world documentation aspect as well.

Learn about data preparation and data cleansing in data science projects to ensure that appropriate data is provided to the next step. Outlier analysis or treatment, handling missing values using imputation, transformation, normalization/standardization, etc., will be explained in thorough detail. Understand the various moments of a business decision and graphical representation so that structured descriptive analytics or descriptive statistics is performed. This exploratory data analytics is the first step in data analytics to draw meaningful insights.

Learn about applying domain knowledge to the data so that more meaningful variables are derived. Understand two main modules of feature engineering including feature extraction and feature selection. Knowing how to shortlist the critical inputs from trivial many inputs is the key to ensuring the high performance of the machine learning models. Understand about extracting features from structured as well as unstructured data such as videos, images, audio, textual files, etc.

Understand one of the key inferential statistical techniques called Hypothesis testing. Understand various parametric hypothesis tests. Learn about the implementation of a Regression method based on the business problems to be solved. Understand about Linear Regression as well as Logistic Regression techniques used to handle continuous as well as discrete output prediction. Evaluation techniques by understanding the measure of Error (RMSE), problems while building a Regression Model like Collinearity, Heteroscedasticity, overfitting, and Underfitting are explained in detail.

Understand the advanced regression models such as Poisson Regression, Negative Binomial Regression, Zero-Inflated models, etc., used to predict the count output variables. Learn about the various scenarios which trigger the application of advanced regression techniques. Understanding and evaluating the models using appropriate performance and accuracy measures of regression are explained in detail.

Data Mining branch called unsupervised learning is extremely important in solving problems, which require the application of only unsupervised learning tasks and also used to support predictive modeling. Clustering or segmentation has two prime techniques – K-Means clustering, as well as Hierarchical clustering and both, are explained in finer detail. Alongside, participants will also learn about handling datasets with large variables using dimension reduction techniques such as Principal Component Analysis or PCA. Finally one will learn about Association rules also called affinity analysis or market basket analysis or relationship mining.

The majority of unstructured data is in textual format and analyzing such data requires special techniques such as text mining or also called as text analytics. Techniques such as DTM/TDM using Term Frequency, Inverse Document Frequency, etc. are explained in this module. One will also learn about generating a word cloud, performing sentiment analysis, etc. Also, advanced Natural Language Processing techniques such as LDA, topic mining, etc., are explained using practical use cases. Also, the learning includes extracting unstructured data from social media as well as varied websites.

A major branch of study in data science is Machine Learning also called Data Mining Supervised Learning or Predictive Modelling. One will learn about K Nearest Neighbors (KNN), Decision Tree (Boosting), Random Forest (Bagging), Stacking, Ensemble models and Naïve Bayes. One will learn about the various regularization techniques as well as understand how to evaluate for overfitting (variance) and underfitting (bias). All these are explained using industry relevant use cases and mini-projects.

Black box machine learning algorithms are extremely important in the field of machine learning. While there is no interpretation in the models, accuracy is unmatched in comparison to other shallow machine learning algorithms. Learn about the Perceptron algorithm and Multi-layered Perceptron algorithm or MLP. Understand about Kernel tricks used within Support Vector Machine algorithms. Understand about linearly separable boundaries as well as non-linear boundaries and now to solve these using Deep learning algorithms.

Understand the difference between cross-sectional data versus time series data. Search about the forecasting strategy employed in solving business problems. Understand various forecasting components such as Level, Trend, Seasonality & Noise. Also, learn about various error functions and which one is the best given a business scenario. Finally, build various forecasting models ranging from linear to exponential to additive seasonality to multiplicative seasonality.

Understand the evolution of AI and Deep Learning and learn the various applications of Deep Learning in building Artificial Intelligence applications. A brief history of Deep Learning and the pace of progress in the space of deep learning is pivotal for budding and emerging data scientists as well as AI experts. Challenges faced in deep learning along with the best practices to overcome the challenges is also explained in detail.

While there are a lot of statistical software and programming languages to perform deep learning activities, Python stands out from the rest. There are a lot of deep learning libraries such as Keras, TensorFlow, Theano, PyTorch, etc., and one will learn about Keras as well as TensorFlow as part of the training module. Image processing is an amazing field to become proficient at and hence you will also learn OpenCV, which stands for Open Computer Vision. The future belongs to Open-source libraries and the fastest development on emerging algorithms will happen in this space. Learning these concepts will help us gain an edge over competitors.

Understanding the treatment of both linearly separable boundaries as well as non-linear boundaries is pivotal for the success of AI experts as well as Data Scientists. In this module, one will learn about handling linear boundaries using the Perceptron algorithm. Understand how weights are assigned and how they are updated each time to reduce the error function. Learn about the Backpropagation algorithm and its application in reducing error using the Perceptron algorithm.

Artificial Neural Network, also called MLP or Multilayer Perceptron is used to handle nonlinear problems. Understand the various network architectures along with different layers including input layers, hidden layers, output layers, etc. Also learn about the various activation functions, error functions, optimization algorithms including Batch Gradient Descent, Stochastic Gradient Descent, Mini-batch SGD, etc.

Understand working with videos and images because the amount of data getting generated in this space is outstripping the volume of textual data. Understand the various features to be extracted from images including edges, textures, etc., by applying various kinds of filters such as Sobel, Harris Corner Detector. Also, learn about face detection using Viola-Jones and tracking human faces in videos. Alongside this also learn about a few image-related models such as image segmentation, image recognition, etc.

Understand how to work with images and videos for building predictive models. Learn about convolution layers as well as handling very small datasets. Understand how to improve the accuracy of models by performing data augmentation activities. Also one should be aware of the use of pre-trained models using feature extraction, fine-tuning, etc., in solving business problems. Finally visualizing the activation layers and heat maps for activation will complete the study to the fullest.

Understand working with textual sequence data and how to perform a one-hot encoding of words and characters. Learn about bi-directional RNNs as well as deep bi-directional RNNs. Learn about various RNN topologies and network architectures. Vanishing and exploding gradient problems are very prevalent in the field of recurrent neural networks. Understand Backpropagation Through Time, which is a different but slight variation from the regular backpropagation algorithm.

Advanced techniques in handling textual and sequential data are LSTMs and GRUs. Also, understand about forecasting temperature. Learn about bi-directional LSTMs and deep bi-directional LSTMs . Also, understand the stacking of various recurrent layers. Stacking recurrent layers will improve accuracy and understanding the same is extremely pivotal for the success of AI algorithms. Also, learn about 1D convolution for time series data. Finally combining CNN and RNN models is an art, which is explained in detail.