Master of Data Science (Global) Program

The act of monitoring, controlling, and analysing marketing performance to maximise efficiency and optimise return on investment is called marketing analytics. Retail analytics examines a variety of business metrics, including revenue, consumer demand, inventory levels, and supply chain activity, among others. In this session, you will learn how to use marketing and retail analytics techniques to analyse a company's performance.

Data visualisation, predictive analytics, information management and segmentation are all used in customer analytics to examine how customers influence essential business choices.

The Konstanz Information Miner, known as KNIME, is an open-source platform for data analytics, reporting, and integration that includes several data mining and machine learning components as well as tools for data pretreatment. In this topic, you will learn how to generate and expand data science productivity using KNIME software, allowing you to concentrate on what you do best.

The Retail Dashboard is a tool for monitoring your company's sales and marketing performance. You will learn how to use the tool to track insights in this module.

The percentage of customers that discontinued using a company's goods and services over a predetermined time is known as customer churn. You will learn how to follow the analysis of clients who discontinued doing business with you in this subject.

A technique used in data mining called Association Rules Mining finds connections between variables in sizable databases. You will learn how to spot patterns among variables from this module.

Credit risk occurs when a borrower doesn't pay back any loans or debts. You will learn how to handle credit risk in this session.

This session will teach you all the critical comparisons of various credit risk models.

The probability that a borrower won't be able to pay back its debts is estimated by Probability of Default (PD) Modelling.

In this session, you will discover more about PD and other models. Additionally, you'll learn how to construct a good model.

Market risk develops when an investor sustains losses as a result of outside forces influencing market pricing. You will study market risk management in this session.

Value at Risk (VaR) is a statistical technique for calculating the financial risk faced by the organisation over a specific time frame.

Here, you will explore the process of identifying patterns of fraudulent behaviour, why fraud occurs and predict where it is likely to occur in the future, ultimately helping businesses detect and prevent fraud.

Data visualisation is the technique of graphically representing data and information. You will learn how to use data visualisation tools in this module to show data trends and patterns.

In this session, you will learn all there is to know about Tableau, the most popular tool for problem-solving using data visualisation.

Here, you will learn how to organise data using charts and a Tableau dashboard.

This session will teach you more about Tableau's descriptive statistics, dimensions and measures.

The science of analytical reasoning, known as "visual analytics", is facilitated by engaging visual user interfaces.

In this session, you will learn about several chart types in Tableau, including line charts, bar charts, pie charts, etc.

You will have your hands full in this session as you implement a case study using Tableau.

You will learn how to integrate Tableau on Google Sheets in this session.

One of the most often used algorithms for predictive analysis in Machine Learning, Linear Regression yields the best results. It is a method that counts on the independent and dependent variables having a linear relationship.

A Supervised Machine Learning approach called Multiple Linear Regression involves many data variables used for data analysis. One dependent variable is predicted using a number of independent variables. You will be guided through all the principles of Multiple Linear Regression that are utilised in Machine Learning.

Like Linear Regression, one of the most widely used ML algorithms is Logistic Regression. It is a straightforward classification technique that can predict categorical dependent variables with the aid of independent factors. In this subject, you will be guided through all the principles of Logistic Regression that are utilised in Machine Learning.

Using Bayes Theorem, classification issues are solved using the Naive Bayes algorithm. In this session, you will learn about the theorem and how to use it to solve issues.

Another well-liked machine learning approach for classification and regression issues is the Support Vector Machine (SVM). Through this session, you will learn how to apply this algorithm.

A supervised machine learning technique called a Decision Tree is applied to classification and regression problems. A Decision Tree is a hierarchical structure where internal nodes represent the dataset features, branches stand in for the decision rules, and each leaf node represents the outcome.

A well-liked supervised learning algorithm in Machine Learning is Random Forest. It includes numerous decision trees on the various subsets of the provided dataset, as the name suggests. The average is then computed to improve the dataset's prediction capability.

A machine learning meta-algorithm called “Bagging”, which is often referred to as bootstrap aggregation, is used to enhance the accuracy and stability of machine learning algorithms used in statistical classification and regression.

A machine learning meta-algorithm under the name of "Boosting" creates robust classifiers from a collection of weak classifiers. Gradient boosting, ADA boosting, and adaptive boosting are other categories of Boosting, which you will understand in this topic.

A well-liked unsupervised learning approach for solving clustering issues in Machine Learning or Data Science is k-means clustering. You will study the algorithm's operation in this topic and then put it into practice.

An ML algorithm called hierarchical clustering creates a hierarchy or tree-like structure of clusters. You can use it, for instance, to cluster several unlabeled datasets into one group inside the hierarchical structure. You will learn how to use and implement this algorithm in this topic.

A method to lessen the complexity of a model, such as reducing the number of input variables for a predictive model to prevent overfitting, is Principal Component Analysis (PCA) for Dimension Reduction. In this topic, you will learn several essentials about the widely used Dimension Reduction-PCA approach in Machine Learning.

Data is transformed through Feature Engineering from its raw condition to one suited for modelling. It transforms the data columns into elements that more clearly depict a particular circumstance. The component's ability to accurately represent an object impacts how well the model can forecast its behaviour. You will learn about several Feature Engineering essentials in this subject.

By comparing each model to the requirements, this subject will show you which one best fits architecture.

You will learn how to use model assessment metrics to optimise the performance of your Machine Learning model in this session.

In this session, you will explore the process of avoiding overfitting and improving model interpretability.

In this subject, you will learn how to use the ML Pipeline to automate Machine Learning workflows. The ML Pipeline can be used by allowing a set of data to be modified and connected in a model that can be tested and evaluated to produce a positive or negative outcome.

By analysing a dataset with replacement, you will gain familiarity with the Machine Learning technique known as "Bootstrap Sampling" to estimate population statistics.

To find the optimal values for any Machine Learning model, hyperparameter tuning is done using the GridSearchCV method. The significance of hyperparameters has a substantial impact on a model's performance. Manually carrying out this procedure is a laborious task. To automate the tuning of hyperparameters, we use GridSearchCV.

Similar to GridSearchCV, RandomizedSearchCV is used to automate the tuning of hyperparameters. For random searches, RandomizedSearchCV is offered, and for grid searches, GridSearchCV is offered.

In Machine Learning, the holdout approach can be improved with k-fold cross-validation. This approach ensures that the performance of our model is independent of the selection of the train and test sets. The holdout approach is applied to each subset of the data set k times, following the division of the data set into k subsets.

As their name implies, Recommendation Systems assist users by predicting their future preferences for certain products and presenting the most appropriate options. You will discover how to employ these methods to guide customers toward the finest products.

Here, you will gain familiarity with the Popularity-based Model, a type of recommendation system that is based on popularity or anything that is currently trending.

First, we explicitly or indirectly gather data from the user. Using this information, we later develop a user profile that will be used to make suggestions to the user. The user increases the system's accuracy by giving us more data or by acting on the recommendation more frequently. The term "Content-based Recommendation System" refers to this technique which we will go through in this topic.

In Collaborative Filtering, many methods are used to find users or products similar to one another in order to provide the best recommendations.

Multiple clustering and classification methods are combined to create a hybrid model. In this section, you will discover how to use a hybrid model.

Here, you will discover how to convert words into numerical vectors with the aid of a technique in Machine Learning called Word Vectorizer. Word Vectorizer allows for applying mathematical and statistical techniques to text data. The technique can be used to improve the performance of Machine Learning algorithms and reduce the dimensionality of the data.

The term frequency of a word in a document is referred to as TF. The simplest method of determining this frequency is to simply count the number of times a word appears in a document.
IDF is a word’s overall inverse document frequency (IDF) over a sample of documents. This conveys the frequency or rarity of a word throughout the entire set of documents. The word is increasingly prevalent when it is closer to 0.

1 Project

Dive into the development stack of ChatGPT by learning the mathematical fundamentals that underlie generative AI. Further, learn about transformer models and how they are used in generative AI for natural language.

Here, you will be introduced to Perceptron, an artificial neuron that is essentially a mathematical simulation of a biological neuron.

Here, you will discover all the uses of Neural Networks, a computing system based on the biological neural network that constitutes the human brain.

A neural network's output is determined by its Activation Function, which takes into account numerous inputs, and the Loss Function is a method for reducing neural network prediction errors. Here, you will explore the application of Activation and Loss Functions.

Finding a function's minima involves an iterative technique called Gradient Descent. Finding the parameters or coefficients of a function's minimum value is done using an optimisation technique. This function occasionally fails to locate a global minimum and can become stuck at a local minimum. In this session, you will discover all there is to know about Gradient Descent.

Here, you will explore the process of Normalisation that involves converting the values of the dataset's numeric columns to a standard scale without distorting the discrepancies between the ranges of values. In Deep Learning, Normalisation is carried out continuously throughout the network as opposed to just once at the beginning. This technique is called Batch Normalisation. A layer's activation function output is normalised before being supplied as input to the following layer.

Google developed TensorFlow, an open-source library for complex machine learning and numerical computation. A powerful open-source API for creating and analysing deep learning models is called Keras. This session will teach you how to execute TensorFlow and Keras from scratch. In Python, these libraries are frequently used for AI and ML.

This session will guide you through every idea related to hyperparameter tuning, an AI training-provided automatic model enhancer.

In this session, you will learn all you need to know about Convolutional Neural Networks (CNN), which are utilised for a variety of tasks, including image segmentation, classification, and processing.

In this session, you will learn how to process an image and extract its data so that it may be used for deep learning image recognition.

In this session, you will discover several techniques in CNN, such as Convolution, Pooling and Padding, and learn about their mechanisms.

The term “Forward Propagation” refers to the process of passing input values through the layers of a neural network until they reach the output layer. The term “Back Propagation” refers to the process of adjusting the weights of the neurons in the hidden layers according to the error in the output layer. These two processes are the basic calculations that are performed to train a neural network.

Here, you will quickly go through the details of popular CNN architectures, such as AlexNet, VGGNet, InceptionNet and ResNet, and how they vary from each other.

Here, you will comprehend Transfer Learning, a research problem in Deep Learning concerned with the storage of knowledge obtained while training one model and its application to another model.

A software program may identify and track down things in an image or video using the Computer Vision approach known as Object Detection. One instance of Object Detection is Face Recognition. This subject teaches you how to use Deep Learning methods to detect any object.

In this subject, you will learn how to detect objects using various Deep Learning algorithms like YOLO, R-CNN and SSD.

In this subject, you will understand semantic segmentation, often referred to as dense prediction in Computer Vision, which aims to assign each pixel of the input image to the appropriate class that corresponds to a particular object or body.

In this subject, you will understand U-Net, a Deep Learning architecture that has been used for image segmentation tasks in Computer Vision. The architecture is based on the encoder-decoder architecture used in convolutional neural networks.

A Siamese Neural Network, also known as a Twin Neural Network, is an artificial neural network that consists of two or more subnetworks that are similar to one another in terms of configuration, parameters and weights. This session will assist you in recognising a person’s face in an image using Siamese Networks.

The goal of object instance segmentation, which advances semantic segmentation, is to differentiate many objects from a single class. Here, you will understand instance segmentation, which is regarded as a hybrid task that combines semantic segmentation and object detection.

By utilising computational linguistics, Natural Language Processing (NLP) creates real-world applications for languages with a variety of structural complexities. With appropriate, effective algorithms, we aim to educate the computer on how to learn languages and then expect it to understand them. The introduction to NLP and all the key ideas you need to understand are covered in this subject.

Stop words do not add anything to the meaning of the sentence and can be eliminated from the sentence without altering its meaning. These words are typically function words such as prepositions, articles and conjunctions. Here, you will discover how to remove stop words in order to reduce the size of the text data set, which in turn, makes processing faster and easier.

Tokenization is the process of breaking down a string of text into smaller pieces called tokens. Tokens can be words, numbers, punctuation marks, or other pieces of text. In this session, you will get familiar with the Tokenization technique used in NLP to split a text into meaningful units that can be analysed.

This session will cover two popular methods in NLP, Stemming and Lemmatization, which make text processing easier and faster by reducing the number of unique words in a text.

Bag of Words is a text modelling method used in Natural Language Processing. To put it technically, it is a method for feature extraction from text data. This method of extracting features from documents is simple and adaptable. In this session, you will learn to keep track of words, ignore grammatical subtleties, word order, etc.

In elementary school, we were taught the distinctions between the many parts of speech tags, such as nouns, verbs, adjectives, and adverbs. POS tagging or POS annotation is the process of assigning each word in a phrase to its appropriate POS (part of speech). Word classes, morphological classes, and lexical tags are other names for POS tags.

Named Entity Recognition, or NER for short, is a standard NLP problem that deals with information extraction. The principal objective is to find and arrange named entities in text into predetermined categories, including names of people, organisations, locations, and events, as well as expressions of time, quantity, monetary values, percentages, and other terms.

A sequence is an organised group of several elements, as the name would imply. This session will teach you how to use the NLP Sequential model to forecast which letter or word will occur.

Recurrent Neural Networks are artificial neural networks that utilise sequential or time-series data. It can be used for speech recognition, image captioning, language translation, and natural language processing. In this session, you will learn about RNNs and discover various mechanisms of RNNs.

This session will familiarise you with the process of vanishing and exploding gradients in RNNs

Here, you will discover how to recognise order dependence in sequence prediction issues using an artificial recurrent neural network called LSTM.

In this session, you will learn about the gating mechanism, called GRU, in RNNs.

In this session, you will learn about all the critical applications of LSTM.

Methods for analysing time-series data to get meaningful statistics and other pertinent data are included in Time Series Analysis. This session will teach you how to forecast future values based on previously observed values with the aid of Time Series Forecasting.

In this session, you will understand how to emphasise the most relevant information when making predictions with the aid of the Attention Mechanism.

In this session, you will comprehend Neural Machine Translation (NMT), a task for machine translation that automatically translates source text from one language to another.

Other popular and sophisticated language models used in NLP will be covered in this session.

We require technical support to make life easier, increase productivity, and make wiser business decisions. We require sophisticated machines in order to accomplish these tasks. While writing programs for straightforward tasks is simple, there needs to be a means to create machines that can handle more complicated jobs. Making devices capable of self-learning is necessary, and here comes the Reinforcement Learning (RL) Framework into the picture.

This session will cover various components of the RL framework.

This session will cover various examples of RL systems.

This session will cover different kinds of RL systems.

In Q-learning, the "Q" stands for quality. It is an off-policy RL system that constantly looks for the optimum course of action given the present situation.

This topic covers everything related to the introduction of GANs (Generative Adversarial Networks), which are deep generative models.

GANs employ a differential function represented by a neural network called a Generator Network, like the majority of generative models.

Another neural network called the Adversarial Network is also a component of GANs, which you will go through in this topic.

In this topic, you will discover how GANs work in Deep Learning.

In this topic, an example will be used to demonstrate how to use Deep Convolutional GANs, which can function as both a Generator and a Discriminator.

You will discover all the necessary and practical applications of GANs in this topic.