COSC2670 Practical Data Science with Python
The Complete Study & Assessment Guide · S2 2026COSC2670 Overview
- RMIT University School of Computing Technologies
- Semester 2, 2026
- a postgraduate course
- 12 credit points
- a practical data-science course using Python
COSC2670 Practical Data Science with Python develops data acquisition, curation, summarisation, visualisation, modelling, classification, clustering, recommender systems and reproducible project practice. It is taught within RMIT University School of Computing Technologies. It is a postgraduate course. It carries 12 credit points.
- Four individual tasks The Handbook publishes a 10/25/35/30 split across quiz, written work and two prototyping tasks.
- No published exam row The accessible 2026 Handbook uses continuous assessment; offering-specific conditions remain controlled by Part B or Canvas.
- Validation before score Keep training, model selection and final evaluation separate so performance is not measured on data used to tune the model.
- Reproducibility is a product A useful result includes data provenance, code, environment, feature decisions and a documented evaluation path.
How COSC2670 is assessed
| Component | Weight | Format |
|---|---|---|
| Quiz or Test | 10% | Individual; detailed timing and conditions must be confirmed in the live course guide |
| Written Work | 25% | Individual data-science written work |
| Design Development and Prototyping 1 | 35% | Individual project work aligned to the published learning outcomes |
| Design Development and Prototyping 2 | 30% | Individual project work; use the live brief for conditions |
The accessible 2026 Handbook publishes four individual tasks worth 10%, 25%, 35% and 30%. It publishes no final-exam row. Course Guide Part B is unavailable, so detailed task and hurdle conditions must be confirmed live.
Current COSC2670 dates
| Date | Item | Control |
|---|---|---|
| 24-30 August 2026 | Canvas Assignment 1 due week | The syllabus does not safely map this label to a Handbook assessment number. |
| 28 September-4 October 2026 | Canvas Assignment 2 due week | Assignment 3 is released in the same week. |
| 19-25 October 2026 | Canvas Assignment 3 due week | Confirm the exact deadline and task mapping in Canvas. |
Current-offering dates captured in the course materials. Confirm changes and exact submission settings in the live LMS.
What COSC2670 covers
The sequence opens at Data Science Lifecycle and Reproducible Python, develops its central analytical shift in Model Design Training and Validation, and closes with Experimental Method and Data Science Reporting.
Data Science Lifecycle and Reproducible Python
data-science lifecycle · notebook · reproducibility · build a reproducible Python workflow from declared inputs02Data Acquisition Curation and Integration
data provenance · data curation · schema integration · combine messy sources while preserving provenance, types, missingness and semantic meaning03Data Summarisation and Visual Evidence
distribution · aggregation · visual encoding · select summaries and charts that expose distribution, group structure and anomalies without hiding the observation level04Model Design Training and Validation
feature · training set · validation strategy · choose features and a train-validation-test structure matched to dependence, clustering and the real prediction setting05Classification Neighbours and Decision Trees
classification · nearest-neighbour rule · decision tree · compare a distance-based classifier with a tree while controlling feature scale, depth and validation06Clustering and Recommender Systems
clustering · similarity · contextual recommendation · select a representation and evaluation for clusters or recommendations without treating an algorithmic grouping as a natural category07Experimental Method and Data Science Reporting
experimental methodology · metric · technical report · turn model output into a report whose claim can be traced to data, method, metric and validation designIt is positioned as a practical data-science course using Python.
The course follows the data-analysis lifecycle through hands-on Python work and evaluates it through four individual tasks rather than a final examination.
Assessment in COSC2670 is distributed as follows: 10% individual quiz or test, 25% individual written work, 35% individual design development and prototyping, and 30% individual design development and prototyping
The operational assessment conditions matter here.
No final-exam row appears in the accessible 2026 Handbook assessment table.
The current Canvas syllabus schedules three assignments across the semester, but it does not safely map every Canvas label to the four Handbook assessment numbers.
What makes COSC2670 demanding is concrete: Designing a reproducible analysis rather than a sequence of library calls: data provenance, curation, feature choices, validation split, metric and interpretation must remain connected to the original problem.
Treat the component-level hurdle status as unconfirmed: the accessible Handbook contains no hurdle statement, while the offering-specific Course Guide Part B is currently unavailable.
For enrolment planning, The 2026 RMIT Handbook publishes no assumed knowledge; students should still confirm program and enrolment rules.
The sequence opens at Data Science Lifecycle and Reproducible Python, develops its central analytical shift in Model Design Training and Validation, and closes with Experimental Method and Data Science Reporting.
Coverage note: RMIT Course Guide Part B is currently unavailable.
Hurdle and detailed task conditions are therefore unconfirmed beyond the accessible 2026 Handbook and current Canvas schedule.
Audit a classification result before accepting its accuracy
- 1Count correct predictions: 72 plus 2 equals 74.
- 1Compute accuracy as 74 divided by 100, giving 74%.
- 1Compute minority recall as 2 divided by 20, giving 10%.
- 1Explain that the aggregate score is dominated by the majority class.
- 1Select a class-sensitive metric and validation split matched to the real decision cost.
Key terms
- Data wrangling
- The process of transforming raw, inconsistent or fragmented data into a documented form suitable for analysis.
- Classification
- Supervised learning that assigns observations to predefined outcome categories from labelled examples.
- Recommender system
- A model that ranks or selects items for a user or context using observed interactions, content or related signals.
- Exploratory data analysis
- Iterative use of summaries and visualisations to understand distributions, relationships, anomalies and data-quality limits.
- Feature
- A measured or constructed input variable used by a statistical or machine-learning model.
- Training set
- The portion of data used to estimate model parameters or learn decision structure.
- Validation set
- Data withheld from fitting and used to compare choices or tune a model before final evaluation.
- Confusion matrix
- A table comparing predicted and actual classes so different kinds of classification error remain visible.
- Clustering
- Unsupervised grouping of observations according to a defined representation and similarity or distance rule.
COSC2670 FAQ
Which current COSC2670 dates are captured?
Canvas Assignment 1 due week: 24-30 August 2026; Canvas Assignment 2 due week: 28 September-4 October 2026; Canvas Assignment 3 due week: 19-25 October 2026. Confirm any change and the exact submission setting in the live LMS.
How is COSC2670 assessed?
10% individual quiz or test, 25% individual written work, 35% individual design development and prototyping, and 30% individual design development and prototyping
Where do students usually lose marks in COSC2670?
Designing a reproducible analysis rather than a sequence of library calls: data provenance, curation, feature choices, validation split, metric and interpretation must remain connected to the original problem.
What is the COSC2670 final assessed-task format?
No final-exam row appears in the accessible 2026 Handbook assessment table. The current Canvas syllabus schedules three assignments across the semester, but it does not safely map every Canvas label to the four Handbook assessment numbers.
What prerequisites or restrictions apply to COSC2670?
The 2026 RMIT Handbook publishes no assumed knowledge; students should still confirm program and enrolment rules.
Is this COSC2670 resource an official university guide?
No. It is an independent COSC2670 study resource; current institutional instructions remain authoritative for assessment operation.
How to prepare for the assessments
Retrieve the course map, practise the recurring method—define the data-science question, audit and transform the data, choose features and a model, separate training from validation, evaluate with decision-relevant metrics and document a reproducible interpretation—on changed scenarios, and verify every operational assessment detail in the live institutional system.
Your AI Data Science tutor for COSC2670
Stuck on a hard COSC2670 question? Sia is AskSia’s AI Data Science tutor — ask any COSC2670 Practical Data Science with Python question and get a clear, step-by-step explanation grounded in how the course is actually taught and assessed. Read this whole study guide free, then take your hardest questions to Sia.
