SEMESTER VIII efficient; Classification and Regression Tree

(CART), Decision Tree modelling.

ICS422 Applied Predictive Analytics [3-0-0-
Textbooks/References
3]
Course Objectives 1. James, Witten, Hastie and Tibshirani “An
● To extend student’s knowledge in the area of Introduction to Statistical Learning:
Data Science with emphasis on Predictions with Applications in R” by, Springer, 1st.
utilizing associated statistical methods and Edition, 2013.
software tools. 2. Alberto Cordoba, “Understanding the
● Students shall better understand the use in its Predictive Analytics Lifecycle”, Wiley,
commercial application to a domain such as 2014.
Telecommunication/ Insurance and should be 3. Eric Siegel, Thomas H. Davenport,
able to independently work on a dataset using “Predictive Analytics: The Power to Predict
RStudio. Who Will Click, Buy, Lie, or Die”, Wiley,
2013.
Course Outcomes 4. James R Evans, “Business Analytics –
● Ability to apply specific statistical and Methods, Models and Decisions”, Pearson
regression analysis methods applicable to 2013.
predictive analytics to identify new trends and 5. R. N. Prasad, Seema Acharya,
patterns, uncover relationships, create “Fundamentals of Business Analytics”,
forecasts, predict likelihoods, and test Wiley, 2015.
predictive hypotheses.
● Ability to develop and use various ICS423 Internet of Things [3-0-2-4]
quantitative and classification predictive
models based on various regression and Course Prerequisites
decision tree methods. The participants should have prior knowledge on
● Develop familiarity with popular tools and the following topics/courses:
software used in industry for predictive ● Computer Networks
analytics, especially R, R Studio. ● Digital Design and Electric Circuits.
● Learn how to apply the models to interpret ● Fundaments of Programming
and report on results for a management
audience for a domain such as Course Objectives
Telecommunications / Insurance. ● To introduce the concept and the basics of
IoT technologies.
Syllabus ● To provide knowledge on various
Introduction to supervised, unsupervised and applications of IoT based technologies and
semi-supervised modelling / learning their associated circuits.
● To enable awareness on the different
Regression model building framework: Problem products that were designed based on IoT.
definition, Data pre-processing; Model building;
Diagnostics and validation Course Outcomes
The students will have
Simple Linear Regression: Coefficient of ● Knowledge on various applications of IoT.
determination, Significance tests, Residual ● Practical knowledge on operating with
analysis, Confidence and Prediction intervals sensors/actuators.
● The capacity to design their own IoT based
Multiple Linear Regression: Multiple coefficient
applications using Arduino or Raspberry PI
of determination, Interpretation of regression
boards.
coefficients, Categorical variables,
Heteroscedasticity, Multi-collinearity, outliers, Syllabus
Auto regression and transformation of variables, Introduction - Sensor basics, sensing and
Regression model building. actuation, basics of networking - wired, wireless,
MANET, PAN, wireless and wired protocols.
Logistic and Multinomial Regression: Logistic
function, Estimation of probability using logistic Communication protocols - IEEE standards, 5G
regression, Deviance, Wald Test, Hosmer era, sensor communications, connectivity
Lemshow Test, Classification table, Gini co- challenges, fading and attenuations.
IIIT Kottayam IIITK/Acad/BCS/Cur./Ver.VI/Senate 13.9(3)/June2024 46/59
IoT architectures and programming - basic Syllabus
architectures, Data processing mechanisms, Introduction: Biological Neuron, Idea of
scalability issues, visualization issues, analytics computational units, McCulloch–Pitts unit and
basics, utility of cloud computing, fog Thresholding logic, Linear Perceptron,
computing, and edge computing, advanced IoT Perceptron Learning Algorithm, Linear
architectures Raspberry Pi and Arduino separability. Convergence theorem for
programming - Golang and Nodejs Perceptron Learning Algorithm.
programming.
Feedforward Networks: Multilayer Perceptron,
Applications - IoT for societal upliftment, Gradient Descent, Backpropagation, Empirical
industrial automation (Industry 4.0), smart city, Risk Minimization, regularization.
smart home, smart transportation, smart
healthcare, smart agricultures, golang based Deep Neural Networks: Difficulty of training
implementation. deep neural networks, Greedy layerwise
training.
Textbooks/ References
1. Pethuru Raj and Anupama C. Raman, The Better Training of Neural Networks: Newer
Internet of Things: Enabling Technologies, optimization methods for neural networks
Platforms, and Use Cases, CRC Press, First (Adagrad, adadelta, rmsprop, adam, NAG),
edition, 2017. second order methods for training, Saddle point
2. Honbu Zhou, The Internet of Things in the problem in neural networks, Regularization
Cloud: A Middleware Perspective, CRC methods (dropout, drop connect, batch
press, First edition, 2012. normalization).
3. Arshdeep Bahga and Vijay Madisetti, Internet
Convolutional Neural Networks: Architectures,
of Things: A Hands-on Approach,
convolution / pooling layers , LeNet, AlexNet.
Universities Press, First edition, 2014.
4. Mung Chiang, Bharath Balasubramanian, Recurrent Neural Networks: Back propagation
Flavio Bonomi, Fog for 5G and IoT through time, Long Short Term Memory, Gated
(Information and Communication Recurrent Units, Bidirectional LSTMs,
Technology Series, Wiley series, First Bidirectional RNNs.
edition, 2017.
5. Alan A. A. Donovan, Brian W. Kernighan, Generative models: Restrictive Boltzmann
The Go Programming Language, Addison- Machines (RBMs), Introduction to MCMC and
Wesley Professional Computing Series, First Gibbs Sampling, gradient computations in
edition, 2015. RBMs, Deep Boltzmann Machines.

IOE421 Deep Learning [3-0-0-3] Deep Unsupervised Learning and Recent

Trends: Autoencoders (standard, sparse,
Course Prerequisites denoising, contractive, etc), Variational
● Machine Learning Autoencoders, Generative Adversarial
Networks, Autoencoder and DBM , Multi- task
Course Objectives Deep Learning, Multi-view Deep Learning.
• The objective of this course is to cover the
fundamentals of neural networks as well as Applications of Deep Learning to Computer
some advanced topics such as recurrent Vision:
neural networks, long short-term memory
cells and convolutional neural networks. Image segmentation, object detection, automatic
• The course also requires students to image captioning, Image generation with
implement programming assignments related Generative adversarial networks, video to text
to these topics. with LSTM models. Attention models for
computer vision tasks.
Course Outcomes
● Identify the deep learning algorithms which Applications of Deep Learning to NLP:
are more appropriate for various types of
learning tasks in various domains. Introduction to NLP and Vector Space Model of
● Implement deep learning algorithms and Semantics Word Vector Representations:
solve real-world problems. Continuous Skip-Gram Model, Continuous Bag-
of-Words model (CBOW), Glove, Evaluations

IIIT Kottayam IIITK/Acad/BCS/Cur./Ver.VI/Senate 13.9(3)/June2024 47/59