Data

Tools

Indexes

Applications

Experiments

Awesome

An open API service indexing awesome lists of open source software.

https://github.com/johannesvc/data-science-portfolio

A curated portfolio of applied data science projects focused on machine learning, NLP, and social impact.
https://github.com/johannesvc/data-science-portfolio

academic-portfolio data-science deep-learning keras machine-learning media-bias nlp pandas scikit-learn

Last synced: about 2 months ago
JSON representation

A curated portfolio of applied data science projects focused on machine learning, NLP, and social impact.

Awesome Lists containing this project

README 

          
## Data Science Portfolio  

**Johannes Van Cauwenberghe**  

London, UK — MSc Data Science (University of London)



This portfolio presents a selection of applied data science projects across domains including natural language processing, causal inference, urban analytics, and ethical AI. Each project folder contains the codebase (`.ipynb` or `.qmd`), dependencies (`requirements.txt`), and a corresponding PDF or HTML report. The goal is not only to demonstrate technical proficiency, but to highlight *how data can inform public-interest questions*---from sustainability and urban policy to media bias and ethics.



---



## Featured Projects



### Optimising News Recommenders Beyond Accuracy (final project)

**[Optimising Beyond Accuracy: Tuning for Diversity and Novelty in Attention-based News Recommenders](https://github.com/JohannesVC/OptimisingBeyondAccuracy)** 

A deep learning–based recommender system for news articles that balances accuracy with diversity and novelty, using the MIND dataset and attention-based neural architectures. Regularisation terms for Intra-List Diversity and Surprisal are embedded directly in the model’s loss function.

📄 [Read report (ResearchGate)](https://www.researchgate.net/publication/391382172_Optimising_Beyond_Accuracy_Tuning_for_Diversity_and_Novelty_in_Attention-based_News_Recommenders)



---



### Visualising Urban Change  

**[Visualising Urban Change: The Impact of Traffic Calming Measures on Symbolic Capital and Socio-Economic Dynamics in London](./visualising-symbolic-capital-London)**  

A spatial analysis of London’s Low-Traffic Neighbourhoods (LTNs) and their relationship to "symbolic capital" retail (boutiques, bakeries, etc.), using Companies House, OpenStreetMap, income/demographic data from statistical agencies, and Voronoi diagrams.

📄 [Read report (PDF)](./visualising-symbolic-capital-London/report-no-code.pdf)



---



### Sustainable Finance under Uncertainty  

**[Bayesian Investment Analysis](./bayesian-investment-analysis)**  

A decision-support system for green investors using a custom Bayesian Network in `pgmpy`, predicting environmental and financial viability from policy and market factors.  

📄 [Read report (PDF)](./bayesian-investment-analysis/report.pdf)



---



### Political Bias Detection in News Media  

**[Political Bias in News: A Feature-Weighted Classifier](./nlp-political-bias-in-news)**  

Classified news articles as left- or right-leaning using the Log-Odds Ratio with Informative Dirichlet Prior (LOR-IDP). Combined AllSides and NewsCatcher data for a bias-aware pipeline.  

📄 [Read report (PDF)](./nlp-political-bias-in-news/report.pdf)



---



### NLP at Scale with Common Crawl  

**[Mapping Corporate Ethics: Clustering UK Companies’ Values Using NLP](./nlp-common-crawl-corporate-ethics)**  

Clustered mission statements from UK firms scraped from the Common Crawl using PySpark, TF-IDF, and topic modelling.  

📄 [Read report (PDF)](./nlp-common-crawl-corporate-ethics/report.pdf)



---



### Systematic Deep Learning for News Classification  

**[A Systematic Exploration of News Classification Using Neural Networks](./deep-learning-for-news-classification)**  

Benchmarked various deep learning models across different news categories. Emphasised reproducibility and interpretability, with structured hyperparameter tuning and model comparison using Keras 3.8.  

🌐 [Read report (HTML)](https://storage.googleapis.com/data-science-portfolio/deep-learning-for-news-classification/report.html)



---



### Multimodal Bias Classification  

**[Multimodal Architectures for Bias Classification in News](./deep-learning-for-multimodal-bias-classification)**  

Integrated textual and metadata features (e.g. source, region, headline) using hybrid deep learning models. Demonstrated that combining non-textual features improves political bias prediction, especially for under-represented classes.  

🌐 [Read report (HTML)](https://storage.googleapis.com/data-science-portfolio/deep-learning-for-multimodal-bias-classification/report.html)



---



### Quantifying Indifference in UK Media  

**[Quantifying Indifference Towards Palestinian Suffering Across UK News Sources](./quantifying-indifference-towards-palestinian-suffering)**  

Analysed over 4,000 articles from UK news outlets (including 744 scraped from BBC News) using NER, sentiment scoring, and date-level casualty records to uncover disparities in how Palestinian vs. Israeli suffering is reported. Offers an interactive data-driven critique of media framing and moral distance, particularly of the BBC.  

🌐 [Read report (HTML)](https://storage.cloud.google.com/data-science-portfolio/quantifying-indifference-towards-palestinian-suffering/report.html)



---



## Navigation and Reproducibility



- Each folder contains its own `README.md` for full context.

- Environments are reproducible via `requirements.txt`.

- Reports are provided as PDFs or self-contained HTMLs.

- Please note that some datasets are not uploaded due to size or licensing.



---



## 📬 Contact  



Feel free to reach out via [LinkedIn](https://www.linkedin.com/in/johannesvc/) or email (johannes.vc@hotmail.com) for collaborations, freelance data work, or research inquiries.