DSM020 Data Programming in Python Coursework: Exploratory Data Analysis

Programme: MSc Data Science, University of London

Module: Data Programming in Python (DSM020)

Weighting: 30% of the total module grade

Coursework Description

Produce a proposal for a data science project of your choosing, which you intend to pursue in the second half of the module. Define the aims and objectives of your project, and use a range of programming techniques to make sure your data is suitable for analysis.

Use critical and analytical skills to explore your dataset through exploratory data analysis (EDA) and identify the key challenges of working with the data. This is your first substantial data processing pipeline and prepares your data for the final coursework assignment of the module.

Describe the formal approach you have taken, including design decisions along the way, such as how the data was captured or retrieved, summary statistics, and refined discussion that helps form concrete research questions for your second coursework assignment.

Submission: a single Jupyter Notebook and any related scripts or SQL files in a single ZIP archive. The notebook should describe your approach and include all processing used to manipulate, cleanse and sanitise the data, plus visualisations and tables. If your dataset exceeds 10MB, include a working sample.

Example Themes

You do not have to choose one of these; they are examples only.

Deliverables

Design a manageable data science project and acquire the necessary dataset in a usable form. Your notebook should include all acquisition steps, pre-processing and changes to the data, with a clear design rationale throughout. For this exercise you should:

• Acquire and prepare your own dataset, making sure you are allowed to share it and anonymising any sensitive information.
• Collate and/or manipulate the data into a usable format, possibly combining multiple sources to verify integrity and accuracy.
• Create a database or flat-file format where needed, for example for very large datasets or frequent CRUD operations.
• Where working with sensitive data, write Python that produces a realistic dummy dataset with sensible distributions and noise.
• Explain the programming techniques used to prepare the data, including any command-line or SQL programming.
• Outline the idea behind your project (context, significance, expected outcomes) and briefly detail what you intend to do with the data.
• Consider any weaknesses or potential caveats in your approach.

Report Structure

Present your report in a single notebook that includes:

• Introduction and context
• A brief description of the dataset (or a sample output), including how it was obtained
• High-level descriptions of the data, its source and its appropriateness for your path of exploration
• A summary of key findings and insights
• Discussion and critical analysis
• Conclusion and further work
• References to any resources used

Visualisations can be shown inline or exported separately (e.g. PNGs) if too large. Include a working sample of your dataset (maximum 10MB) and a requirements.txt file with instructions on how to replicate your approach.

Marking Rubric (10 marks each)