Queen's School of Computing

CISC351/3.0 Advanced Data Analytics

Calendar description:

Design and implementation of complex analytics techniques; predictive algorithms at scale; deep learning; clustering at scale; advanced matrix decompositions, analytics in the Web, collaborative filtering; social network analysis; applications in specialized domains.
Prerequisites: (APSC 42 or APSC 143 or CISC 101 or CISC 110 or CISC 151 or CISC 121 or previous programming experience) and CISC251 and (STAT263 or STAT options)
Learning Hours: 120 (36L;24Lab;60P)

Course Outline:

Advanved Analytic Algorithms (6 weeks)

  • Deep learning (restricted Boltzmann machines, stacked denoising autoencoders, prediction and clustering using deep networks, the adversarial example problem)
  • Gradient boosting trees
  • Generalized linear models
  • Scaling clustering algorithms (DBNorm)
  • Advanced clustering using matrix decompositions (Singular Value Decomposition, Independent Component Analysis, Semi-Discrete Decomposition, Non-Negative Matrix Factorization)
Application to non-tabular data (4 weeks)
  • Web analytics (PageRank, HITS, reducibility problem)
  • Collaborative filtering and recommender systems
  • Social network analysis (social network properties, viral spread, spectral embedding)
  • Text analytics
Scaling (1 week)
  • Scaling algorithms to larger data, introduction to clusters for analytics
Examples (1 week)
  • Examples of real-world applications in natural language analysis, security, marketing

Learning outcomes:

Upon successful completion of this course, a student will be able to:

  1. Design inductive model building algorithms appropriate for datasets of substantial size and complexity with ill-defined requirements
  2. Plan ways to collect data, build models, and interpret results in network datasets
  3. Evaluate the modelling performance of such algorithms, and the implications for the real-world system that the data describes

Possible Textbooks:

  • Zaki and Meira, Data Mining and Analysis: Fundamental Concepts and Algorithms, Cambridge University Press.
  • Skillicorn, Understanding Complex Datasets, Taylor and Francis