GEOG0178 Machine Learning for Social Sciences with Python
COURSEWORK INSTRUCTIONS
COURSEWORK: Machine learning-data science project
Deadline – noon 28 April, 2025
The objective:
The objective of this coursework is to analyse a dataset of your choice in order to provide a data-based answer to a research question. Thus, naturally, your first tasks include figuring out a research question that interests you and finding a dataset that can help you answer it by applying machine learning modelling to it. Big data producers such as the UN, the World Bank, the OECD, the Eurostat may be of interest in terms of data search. But you are welcome to explore further. Any topic from social sciences can be analysed. The final output of this assessment is a coherent report (supported by a jupyter notebook that contains the python code for the analysis).
Report structure:
Students should submit the report (in a .pdf format) through Turnitin on the course Moodle page, under the “Assessment” tab. The report should be coherent and structured as follows:
1. Introduction
· Background, context and research question
2. Brief literature-based overview of the topic and the research question
· Cite a minimum of 10 academic and/or policy papers
3. Data and method
· Variables
· Why the selected machine learning model(s) is appropriate for the selected dataset/question? Main model premises.
· Discuss data cleaning/wrangling (if applicable)
4. Interpretation and discussion of machine learning modelling results
· Exploratory data analysis (EDA)
· Model results and performance
· Comparison of models***
· Limitations and implications
5. Conclusion
· Summary of the main findings
· Implications of the findings
· How could the analysis/model be improved?
· Suggestions for further research within the topic
***Important note: students are required to run at least two machine learning models in their analysis and compare their performance. Note that the two (or more) models do not necessarily need to be different machine learning techniques. They can be two or more model variations conducted with the same machine learning technique (e.g., random forest with 8 variables as the 1st model and random forest with 5 variables as the 2nd model).
Submission format:
The report should start with the UCL Geography cover page you can download from the link inserted in the box above.
In a PDF document with text of font size 11 or 12 and written fully in complete sentences, e.g. not using bullet points and not including any Python code in the report.
The report’s maximum length is 2,000 words (+/- 15% rule DOES NOT apply) which you are free to divide in any way between the sections and subsections. Please respect the maximum word count of 2,000 words.
The word count includes headings and subheadings, main text but excludes the coursework cover page, title, captions of figures or tables, bibliography (list of references) and appendices (if there are some) at the end of the document.
The maximum number of figures is 10 in total (multiple sub-figures used to make the same point are allowed) and the relevance of these figures should be explained in your write-up.
The code developed by the student should be submitted using a separate submission link available on the course Moodle page in a single. ZIP (compressed) file. The code can be submitted as a Jupyter notebook(s), i.e. .ipynb file but it must be contained within one ZIP file. Please add your data file(s) to the ZIP file.