Course Outline, Winter 2024
Big Data Analytics
SOEN 471 / SOEN 6111

Instructors

Coordinator: Dr. Tristan Glatard
e-mail: tristan.glatard@concordia.ca
Regular (online or onsite) office hours: Wednesday 4pm - 5pm or by appointment. Office: ER 9.919. A Zoom link will be posted on Moodle.

Teaching Assistants:

• Inés Gonzalez Pepe (Projects)
  e-mail: inesgp99@gmail.com
• Mathieu Dugré (Lab assignments)
  e-mail: math.dugre@gmail.com
• Sephora Maltais (Lab assignments #2)
  email: seph156@gmail.com

Lectures & Labs

• Lectures: Wednesday 5:45PM - 8:15PM. H 110.
• Labs:
  • Wednesday 3:45PM - 5:35PM.
    • Room H903 (assignment lab #1, Mathieu)
    • Room H849 (project clinic, Inés)
  • Wednesday 8:30PM - 10:20PM.
    • Room H817 (assignment lab #1, Mathieu)
    • Room H967 (assignment lab #2, Sephora)
    • Room H903 (project clinic, Inés)

Objectives

Big Data analytics has been transforming industry and science in various domains for the past few years, making possible the processing of Terabytes of data on a daily basis. This was enabled by the joint evolution of programming models, data-analysis algorithms and computing infrastructures.

This course introduces the concepts and some of the main algorithms used for Big Data analytics. It presents the principles of the Hadoop ecosystem, Apache Spark, and it details the main algorithms for the analysis of large datasets, related to similarity search, mining of frequent itemsets, graph analysis, clustering, stream mining, recommender systems and advertising.

By the end of this course, students will be able to write and deploy efficient parallel algorithms to analyze Big Data sources for various applications.

Communication

Important information will be communicated through Moodle and/or Slack. Students are expected to consult these channels regularly.

• Moodle page
• Slack workspace (join channel #bigdata--winter2024)

• Spend a few minutes searching for answers on your own before asking other people.
• Keep an eye on the on-going discussions, and try to avoid asking a question that has already been asked.
• Use Slack's thread feature to follow-up on a question instead of posting replies to the main channel.
• Don't hesitate to chime in a discussion if you think you might help.
• Always use the public channel (#bigdata--winter2024) to ask your questions. In particular, never exchange private messages with your TA. If any personal matter has to be discussed, communicate with the professor. If you really need to send a non-public message to your TA, always involve the professor.

The instructors are dedicated to providing a harassment-free experience for everyone, regardless of gender, gender identity and expression, sexual orientation, disability, physical appearance, body size, race, age or religion. We do not tolerate harassment of students or TAs in any form. Sexual language and imagery is not appropriate for any communication, in particular on Slack. For more information, please consult Concordia's policy on harassment.

Schedule

Please note: In the event of extraordinary circumstances beyond the University's control, the content and/or evaluation scheme in this course is subject to change.

Book

• MMDS (Required): Mining of Massive Datasets, Jure Leskovec, Anand Rajaraman, Jeff Ullman, beta version of the 3rd edition. Available online.

Course Evaluation

Lab assignments (15%): You will be required to develop data analysis programs in Python using Apache Spark or Dask. There will be a total of three assignments. You must work on these assignments individually. The lab assignments are all due on a Friday evening, 11:55pm (see exact dates on the schedule table). A grace period of 48 hours will be automatically granted (assignments will be accepted until Sunday night, 11:55pm), but no further extension will be granted. Assignments must be submitted through GitHub Classroom, you will receive a link for each assignment.

Exams (60%): There will be a mid-term and a final exam. All exames will be closed book and no electronic device will be allowed except ENCS calculators. Exams will be Multiple-Choice Questionnaires. The midterm will be conducted in-class and will count for 20% of the final grade. The final exam will count for 40% of the final grade. There will be no substitution for a missed exam.

Project (25%): This course will walk you through the definition and implementation of a data-science project using Big Data technologies. During the project clinics and lecture, the instructors will guide you through the following milestones:

1. Project definition:
   • Option 1: Own Project. You will define your own project by identifying: (1) a dataset of interest, (2) a set of research questions to be answered with the dataset, using techniques studied in class. If you are doing a Master or PhD thesis, you are encouraged to define a project linked to your research topic.
   • Option 2: Industrial Project. You will select a project proposed by a company through the Riipen platform. Industrial projects will be validated by the instructors and published on January 24.
   Both project types are expected to follow the same milestones (described below). Milestones for industrial projects might be slightly revised depending on the exact nature of the projects.
2. Model design: choose a class of models in {supervised learning, recommender system, clustering, frequent itemset}. Outline how the data model could be applied to your dataset to answer your research question. Research algorithms and techniques to implement this class of model.
3. Data preparation: inspect the dataset, identify missing data, outliers, data types (categorical data in particular), and write Spark or Dask programs to correct for potential issues.
4. Model implementation: implement the model with Spark, Dask or scikit-learn.
5. Model evaluation: identify evaluation metrics for your model. Implement and discuss them.

1. Team registration (1%) Easy point! Please register your team of 4 students on time to allow for a smooth organization of project clinics.
2. Participation (3%) Attendance and participation to project clinics. Project clinics will follow the following template:
   • Project team presents updates to class (5')
   • Instructor gives feedback and directions (5'). Audience may ask questions.
   • Feedback on project milestones (3').
   Don't miss an opportunity to get feedback on your project!
3. Project summary (3%) The project summary will be a 400-word abstract available as a Markdown (.md) document in a GitHub repository. The summary will report on project definition and model design. It will describe the dataset and its main characteristics (number and type of features), the research questions to be addressed in the project, the class of models to be applied to the dataset, and the algorithms that will be used. At least two algorithms must be used and compared. The project summary will be evaluated during the project clinics (see schedule table).
4. Project data model (10%) The project data model will be delivered as a Jupyter notebook containing code and explanations to implement data preparation, model training and (optional) model evaluation. The project data model will be evaluated during the project clinics (see schedule table).
5. Final project presentation (8%) The project presentation will be delivered during the last week of the course as a 6-10 minute presentation putting special emphasis on model evaluation and summarizing the other project milestones.

Grading Scheme: There is no standard relationship between percentages and letter grades assigned. The grading of the course will be done based on the relative percentages assigned to the assignments, project and the exam. There is no definite rule for translation of number grades to letter grades.

Academic Integrity