Deep Learning for Crime Forecasting: The Role of Mobility at Fine-grained Spatiotemporal Scales - Journal of Quantitative Criminology Objectives To develop a deep learning framework to evaluate if and how incorporating micro-level mobility features, alongside historical crime and sociodemographic data, enhances predictive performance in crime forecasting at fine-grained spatial and temporal resolutions. Methods We advance the literature on computational methods and crime forecasting by focusing on four U.S. cities (i.e., Baltimore, Chicago, Los Angeles, and Philadelphia). We employ crime incident data obtained from each city’s police department, combined with sociodemographic data from the American Community Survey and human mobility data from Advan, collected from 2019 to 2023. This data is aggregated into grids with equally sized cells of 0.077 sq. miles (0.2 sq. kms) and used to train our deep learning forecasting model, a Convolutional Long Short-Term Memory (ConvLSTM) network, which predicts crime occurrences 12 hours ahead using 14-day and 2-day input sequences. We also compare its performance against three baseline models: logistic regression, random forest, and standard LSTM. Results Incorporating mobility features improves predictive performance, especially when using shorter input sequences. Noteworthy, however, the best results are obtained when both mobility and sociodemographic features are used together, with our deep learning model achieving the highest recall, precision, and F1 score in all four cities, outperforming alternative methods. With this configuration, longer input sequences enhance predictions for violent crimes, while shorter sequences are more effective for property crimes. Conclusion These findings underscore the importance of integrating diverse data sources for spatiotemporal crime forecasting, mobility included. They also highlight the advantages (and limits) of deep learning when dealing with fine-grained spatial and temporal scales.

🚀 I'm very excited to share that my first first-author paper "Deep Learning for Crime Forecasting: The Role of Mobility at Fine-grained Spatiotemporal Scales" is now available in the Journal of Quantitative Criminology!

More below 👇
link.springer.com/article/10.1...

📢 Do you want to join our group? We opened **two** calls for a fully funded PhD. Details are in the image and at the following links.

Calls: iecs.unitn.it/education/ad...
PhD Details: iecs.unitn.it/education/ad...

Deadline: August 22nd, 2025, hrs. 04:00 PM (CEST)

We introduce a real-time method to infer individual unemployment using GPS trajectories and survey data.
By analyzing mobility patterns of ~1 million individuals before and after job loss we reveal a sustained contraction in exploration that deepens with time since job loss.

🚀Job loss disrupts individuals’ mobility and their exploratory patterns🚀
Thanks to a great collaboration with @marcodena.bsky.social, @marcotonin.bsky.social, Bruno Lepri and @lorenzolucchini.bsky.social our latest study is finally out in iScience!

So excited to see this come together! 🎉

Our latest study explores the interplay between science and misinformation in public debates during COVID-19 🔍 arxiv.org/abs/2507.01481

👇Take a look

Latest out in PNAS!! Comparative evaluation of behavioral epidemic models using COVID-19 data. Amazing collaboration with @ngozzi.bsky.social and @alexvespi.bsky.social www.pnas.org/doi/10.1073/...

Forecasting Seasonal Influenza Epidemics with Physics-Informed Neural Networks Accurate epidemic forecasting is critical for informing public health decisions and timely interventions. While Physics-Informed Neural Networks (PINNs) have shown promise in various scientific domain...

Forecasting Seasonal Influenza Epidemics with Physics-Informed Neural Networks arxiv.org/abs/2506.03897

Congrats Giovanni!!

New paper alert. "More is different" & the LLMs.
LLMs are usually studied in isolation. But what happens when they start interacting? We explored this by looking at their collective behaviour.
Work with @ariel-flint.bsky.social and @lajello.bsky.social
1/

We present Epydemix: open-source python package for all stages of epidemic modelling: from models' definition to their calibration via ABC methods. Website: www.epydemix.org. Paper: www.medrxiv.org/content/10.1.... High level summary: www.epistorm.org/activities/e...

Lecturer in Applied Mathematics at City St George’s, University of London An academic position as a Lecturer in Applied Mathematics is being advertised on jobs.ac.uk. Click now to find more details and explore additional academic job opportunities.

*Permanent* position in Applied Mathematics at City. We are - of course - particularly interested in profiles in Computational Social Science, Network Science, Data Science, and related fields. The application deadline is June 1st.

www.jobs.ac.uk/job/DMY889/l...

🎉 New paper out in PNAS! 👇👇👇

Happy to see our work, "Socioeconomic disparities in mobility behavior during the COVID-19 pandemic in developing countries", out in the EPJ Data Science special issue "Data for the Wellbeing of Most Vulnerable".

bit.ly/socioeconomi...

Thanks to the editors and the amazing team for the hard work!

Stylized map of Detroit (MI) showing the highway network, and the network of social connections between urban residents. The connections intersecting highways are sparser than elsewehere. Image credit Karo Berghuber (Insta: @kariot.lines)

"Urban Highways Are Barriers to Social Ties" out on PNAS!
The 1st large-scale measure of how highways weaken social connections between the communities they separate. This barrier effect is strong in the 50 largest US cities--especially for low-income Black communities.
www.pnas.org/doi/10.1073/...

If education is expensive, try ignorance.

Using human mobility data to quantify experienced urban inequalities - Nature Human Behaviour Xu et al. review applications of urban mobility behaviour data and propose a temporal bipartite network that reveals mobility patterns between people and places. It helps to track urban inequalities i...

🎉 Today, our perspective on "Using mobility data to quantify experienced urban inequalities" is finally published in Nature Human Behavior. I am honored to be listed among many stellar coauthors and thankful for their valuable insights.
www.nature.com/articles/s41...

Most tests for LLM biases use questionnaires, asking the model to generate a stance towards a given topic. Sadly, biases can re-emerge when the model is used in the application context. We show that apparently unbiased LLMs exhibit strong biases in conversations.
Preprint: arxiv.org/abs/2501.14844