The Banking Analytics Lab

The Banking Analytics Lab at Western University invites applications for two fully funded PhD positions in credit risk modelling, financial stability, and advanced machine learning). In addition, we invite applications for up to two fully funded PhD positions focusing on climate risk. These research projects aim to develop rigorous and objective methods to improve credit risk management and financial risk assessment. The research will leverage alternative data sources, network-based stress testing frameworks, advanced deep learning methods, large-scale models, and robust optimization techniques. Within this framework, the research will also address how multiple sources of risk, including climate risks, can be incorporated into credit risk and financial stability analysis across economic sectors, with applications in banking, finance, and regulation.

Position Details:

Full funding guaranteed for four years.
Annual stipend of CAD$30,000.
Direct collaboration opportunities with major banks, pension funds, data vendors, and regulatory bodies.
Access to unique datasets and computing resources, plus annual funds for travelling and expenses.

The Successful Candidates Will

Conduct research using advanced deep learning methods and large-scale models to address key challenges in credit risk management, including the assessment of alternative data value, risk propagation and financial contagion in payment networks, individualized pricing strategies, and robust optimization under portfolio level shocks.
Develop deep learning models and network-based stress testing tools to assess how such events impact credit risk and financial stability, using real nationwide microentrepreneur data in a partnership with leading financial institution.
Develop new analytical and computational frameworks that explicitly integrate climate scenario projections related to economic transition and physical shocks into financial risk assessment for institutional investing and regulatory stress testing.
Work at the intersection of applied probability, econometrics, machine learning, statistics, stochastic control and banking using real, challenging datasets to solve problems at the forefront of modern banking and climate risk modelling.
Contribute to both academic research and practical applications.
Engage with industry partners and regulatory bodies.

Required Qualifications:

Master’s degree in statistics, operations research, computer science, economics, applied mathematics, or related fields for PhD positions in banking analytics.
Master’s degree in mathematical finance, statistics, operations research, economics, or related fields for PhD positions in climate risk.
Strong quantitative and programming skills. Python programming and knowledge of modern methods (pytorch, polars, spark, arrow, duckdb) is a strong plus.
Excellent written and verbal communication abilities.
Demonstrated interest in banking, financial, or mathematical finance applications.

Application Process:

Submit your CV, academic transcripts and reference letters by January 31^st, 2026 on Western’s graduate portal: https://grad.uwo.ca/admissions/apply.html. Please mention your GRE scores, if available. More information available here: https://www.uwo.ca/stats/graduate/phd-program.html
Shortlisted candidates will be invited for interviews in February 2026.
Final acceptance will be determined by the Graduate Affairs Committee.

For more information on positions in banking analytics, please contact Cristián Bravo at cbravoro@uwo.ca, and for positions in climate risk, please contact Ankush Agarwal at aagarw93@uwo.ca

One final, co-supervised, position is available in the UK, please see for this opportunity here with The University of Southampton.

The Banking Analytics Lab values diversity and encourages applications from all qualified individuals, including women, members of visible minorities, Indigenous peoples, and persons with disabilities. This position is available to anyone, worldwide.

Our latest paper is live! This is the work of our Ph.D. researcher Sahab Zandi, co-supervised by Cristián and María, in collaboration with Christophe Mues, J. C. Moreno-Paredes, at Santander Bank at the time and also Cristián’s Ph.D. classmate, and Kamesh Korangi. This paper is the conclusion of two years of work on our collaboration with Santander Spain, who provided direct access to their database. Our work is, to the best of our knowledge, the first one that measures calculates how risk propagates on small businesses on an individual basis, using advanced deep learning paired with financial behaviour and transactional information.

The Problem We Set Out to Solve

SME lending is challenging to judge from spreadsheets alone. Many small firms don’t have long financial histories, and traditional ratios can miss how trouble moves through supply chains. If a key customer stops paying, cash dries up and loan payments follow – that risk lives in relationships, not just in financial information.

The Idea

We added a “map of relationships” to the usual credit data. One layer captures who pay whom (financial transactions); another captures who share ownership. Together, those multilayer links give a fuller picture of where stress might spread.

How We Built It

We trained models that look at both worlds at once: the tabular records lenders already keep and graph representations of the two networks. A cross-attention fusion step lets each side inform the other before making a call – so the model isn’t blind to either the firm’s own profile or its neighbourhood.

What Changed

Adding the networks made predictions sharper than using standard data alone, with our best cross-attention model on the double-layer graph topping the baselines. We were able to catch more future defaulters, more reliably. We also saw modest gains when we kept the real-world direction and size of money flows, which help the model tell strong ties from weak ones.

What the Links Actually Tell Us

The transaction layer tends to be more informative than common ownership because it reflects live dependencies: who owes you money and how concentrated those receipts are. Firms exposed to defaulted customers (money flowing in from a troubled partner) look riskier than those exposed to defaulted suppliers, an effect that’s stronger when the ties are big and recurring.

Why It Matters

For lenders, network-aware scores flag clusters of correlated risk earlier and support fairer pricing. For SMEs, diversified, healthy counterparties become a credit asset in their own right. More broadly, better risk signals can widen access to finance—especially where traditional histories are thin—while still keeping decisions explainable and auditable.

Read the preprint in this ArXiV link. The paper is still a preprint, so please take it with a grain of salt as it will evolve through the peer-review process. We welcome any feedback!

It was a pleasure to be back on the CBC News Weekend Business Panel this Saturday. We had a packed session (we even had to leave a few topics out!), diving into the escalating U.S. tariffs on Canadian goods, the historic leveraged buyout of Electronic Arts, and the ongoing labour dispute at Canada Post.

Here’s a summary of my main points on each topic.

Topic 1: The Impasse at Canada Post

The current labour dispute at Canada Post stems from two fundamentally opposed visions for its future. Management, facing a decline in letter mail from 5 billion to 1.5 billion pieces and stiff competition in parcel delivery, wants a leaner, more agile corporation. The union, conversely, is fighting for job security, investment, and benefits. These goals are, at their core, contradictory.

The corporation’s latest offer, which the union called “a step backwards,” clearly reflects management’s vision. It proposes decent exit packages to reduce headcount, the reclassification of rural offices to allow for closures, and no signing bonus. It’s a cost-cutting proposal, plain and simple.

The stark reality is that Canada Post is currently insolvent, running a $1.5 billion annual deficit. This leaves us with a fundamental choice as a country: do we want to heavily subsidize the corporation to maintain its current scope and service levels, or do we accept a significant reduction in its operations? The current leadership is firmly in the second camp, while the union is in the first. With such a wide gap, a resolution does not appear to be on the immediate horizon.

Topic 2: Escalating U.S. Tariffs on Canadian Goods

President Trump’s latest tariff announcement is a significant escalation of a long-standing problem, particularly for our softwood lumber industry. The core change this time is that there are no carve-outs for CUSMA-compliant goods. These new duties are being stacked on top of a 35.16% in combined anti-dumping and countervailing duties.

The U.S. justification is an accusation of “targeted dumping,” which they calculate using a controversial method called “zeroing.” In simple terms, they essentially ignore fairly priced trades to make the impact of any underpriced sales appear much larger than they are. This policy will hit Ontario and British Columbia the hardest, compounding the economic pressure already felt by the auto sector.

Legally, our options are limited. The WTO’s appeal body is non-functional, a holdover from the first Trump administration. While several legal challenges are pending, Canada has recently dropped two appeals, suggesting we may have a losing hand. The most viable path forward seems to be a negotiated settlement, likely involving a fixed quota on Canada’s total market share in the U.S. Separately, the threat of pharmaceutical tariffs would also disproportionately impact New Brunswick, where up to 7% of the province’s GDP is tied to this sector, dragging it into a bruising trade war.

Topic 3: The Historic Leveraged Buyout of Electronic Arts

The $55 billion leveraged buyout (LBO) of Electronic Arts by a Saudi-backed consortium and Silver Lake is the largest in history, and its structure is fascinating. Unlike typical LBOs, which are often 60-90% debt, this deal is financed with approximately 65% equity. While the total debt ($20 billion) is massive, the high equity stake makes this a relatively low-risk LBO from a leverage perspective.

This deal is part of a larger $3 trillion M&A trend this year, spurred by reduced political uncertainty and lower regulatory oversight. For Saudi Arabia’s Public Investment Fund (which already owned 10% of EA), this is a step forward in its economic diversification strategy. From a business standpoint, EA is an ideal target. As gamer myself, I’ve played several of its franchises like Mass Effect and Dragon Age, together with their real workhorses EA Sports and Battlefield, they generate a stable $2 billion in yearly cash flow, which is very attractive for servicing the debt taken on in an LBO.

The future potential is also a huge factor, which would suggest why there is so much equity into the deal. AI can dramatically shorten game development cycles and reduce costs, potentially unlocking enormous future growth for the industry. While LBOs often lead to cost-cutting that can harm quality, EA’s existing reputation and potential future efficiencies gives the new owners some leeway.

If you missed the live discussion, you can watch the full segment below.

Our new book is out for preorder! We have been working hard to get this book out, and we are thrilled to finally share it with you. While the book comes out in December 2025, you can already go through the labs! Find them here. A preliminary Table of Contents, Figures and Algorithms can be found here.

Deep Learning in Banking book cover

The book summarizes close to a decade of experience in applying deep learning to banking problems, with a focus on risk management. We are proud to have written a book structured for both academic and professional use. Our target is data scientists developing models, risk managers evaluating them, regulators shaping policy, and graduate students preparing for a career in financial technology. We assume you have a background in data science, but we guide you through the specific nuances of applying deep learning in the high-stakes environment of banking.

What’s Inside?

We wanted to create a single resource that bridges the gap between cutting-edge theory and the practical realities of banking. The book provides:

A Comprehensive Toolkit: Dive deep into the essential deep learning architectures—from Convolutional Neural Networks (CNNs) for image analysis to Transformers for text and Graph Neural Networks (GNNs) for understanding financial contagion.
Real-World Case Studies: Each chapter is grounded in practical applications, using real data to demonstrate how these models can be applied to solve core banking challenges like mortgage default prediction, behavioral scoring, and risk analysis. We are proud to have received special permission from Freddie Mac to use their Single Family Loan Dataset for our book. We also use Fed speeches, network data, LiDAR, and a long list of alternative data for the book. All case studies are done on real data, with direct application to banking practice.
Focus on Trust and Compliance: The book dedicates significant attention to the critical themes of fairness, accountability, explainability, and the ever-evolving regulatory landscape, including frameworks like the EU AI Act and regulators worldwide.
Hands-On Learning: Every case study is accompanied by language-agnostic algorithms in the book, so it will never go out of fashion, and always up-to-date labs are available on the book’s companion website. This allows you to not just read about the models but to build and experiment with them yourself. You can try these right now!

Our goal was to create the book we wished we had: a single, unified resource that covers the data, the algorithms, and the business environment of AI in financial services. We believe that responsible AI can drive innovation and growth in banking, and this book is designed to give you the tools to build solutions that are safe, profitable, and productive. Interested? Preorder it now!

Our latest paper, from our lab member Mahsa Tavakoli, is out. This was in collaboration with Prof. Rohitash Chandra at UNSW in Australia

When it comes to understanding how credit ratings are determined, most studies have focused only on numbers—like financial ratios and balance sheets. But in the real world, a lot of important information is found in written documents, such as company reports or news articles. In our study, we looked at how combining both numbers and text using deep learning models could improve the prediction of credit ratings. We tested different types of models and ways to blend the data, and we found that a model based on CNN (a type of deep learning model) that mixes information early and in the middle of the process gave the best results. Surprisingly, simpler models worked better than more complex ones, and the text information—like what’s written about a company—was even more useful than numbers in predicting credit ratings.

We also checked how reliable our model is when things change, such as during big events like the COVID-19 pandemic. The results showed our model stayed stable and still made good predictions, especially when using text data. Another interesting finding was that credit ratings given by Moody’s were more accurate over time than those from other rating agencies. This could help financial institutions trust those ratings more when making decisions. Overall, our study shows that using both text and numbers together can lead to smarter and more reliable credit rating predictions. It also opens the door for better tools that can help people make confident financial decisions—even during uncertain times.

The paper was published a bit ago at Applied Soft Computing. You can find the paper Open Access here, and its ArXiV version here.

I was on the CBC News’ Weekend Business Panel this week speaking about Trump’s Tariffs, part 2, and the end of the HST holiday.

No one wins in a trade war. The 2018 aluminum and steel tariffs cost 75,000 jobs in the US and 2,000 people in EI, just in the few months they were up. Trying to choose winners and losers creates a few temporary winners at the cost of everyone else. The BoC is now being more measured, as they know they won’t be able to respond with force if we have tariff-induced stagflation.
The HST holiday was, as I said in the panel in December, a whole lot of nothing. Almost no growth impact, with only restaurant owners and workers reporting a significant increase in sales. We are waiting to see its impact on the government’s debt, which, if meaningful, would mean we are all paying for this in the future.

Give it a watch below!

I started this year’s participation on the CBC News’ Weekend Business Panel by discussing what I’m sure will be common in my future media apparitions: Incoming President Trump’s policies and how they affect Canada.

This week, I discussed the new tariff threat and Canada’s response (they would be severe, and we should not be showing our hand to have a stronger negotiating position) and the TikTok ban (the influencer economy will be fine, but there is much more at stake than simply that, it’s a geopolitical issue).

Give it a watch below!

I was on the CBC News‘ Weekend Business Panel this Saturday, speaking about three topics:

1. The Canada Post Strike: The company is in dismal financial status, with seven years of losses, each year beating the next. The company wants to turn more into an Amazon-like delivery service, with contractors delivering parcels over the weekend, while the union wishes for their members to get paid overtime for these deliveries. Both of these requirements are unrealistic. Canada Post functions in a highly competitive environment, one where labour laws are many times overlooked. A solution must come via rethinking what is as modern post service that reaches rural and urban Canadians, possibly some reforms to the labour code to protect delivery workers across companies and provide a level playing field, and a more lean, efficient post office that delivers services as Canadian need them. Otherwise, the crown corporation is, in my opinion, doomed.

2. The GST holiday and the $200 incentive: This is a terrible policy. At best, it displaces consumption and reduces the fiscal arks with limited economic impact. At worst, it compounds an inflationary environment given the promised increased expenses by the incoming US president. If the latter occurs, then the BoC will react and either stay further rate cuts or, in a more extreme situation, increase the interest rate, eliminating any impact. This is a populist measure that has, sadly, proven quite popular by Doug Ford’s similar measures.

3. The latest inflation numbers: They were squarely in the BoC’s estimates, so nothing serious here. I do believe that the BoC hast to be cautious about the future. Donald Trump’s protectionist and expansionary policies may lead to an even weaker CAD, thus the risk of importing inflation in the future is high. The BoC must be thinking carefully whether they can keep cutting the interest rate or should they wait to see the impact of Trudeau’s, Ford’s and Trump’s policies.

Give it a watch below! This is my last panel of the year. I come back live on January 18th.

I am recruiting up to two new Ph.D. students for entry September 2025. I am running a shorter recruiting cycle this year given the new constraints on international students (Deadline: December 2nd, 2024!). The ad below:

Ph.D. Position in Banking Analytics – Department of Statistical and Actuarial Sciences, Western University

The Banking Analytics Lab at Western University invites applications for a fully funded Ph.D. position focusing on financial contagion in corporate and consumer credit risk. This cutting-edge research project aims to understand and model the interconnected nature of credit risk across different sectors of the economy, with direct applications in banking and financial regulation.

Position Details:
– Full funding guaranteed for four years.
– Annual stipend of CAD$30,000.
– Direct collaboration opportunities with major banks and regulatory bodies.
– Access to unique datasets and computing resources, plus annual funds for travelling and expenses.

The successful candidate will:
– Develop novel methodologies for analyzing financial contagion using deep learning techniques in consumer retail, small business lending, and/or corporate lending.
– Work at the intersection of machine learning, statistics, and banking over real, challenging datasets solving problems at the forefront of modern banking.
– Contribute to both academic research and practical applications.
– Engage with industry partners and regulatory bodies.

Required Qualifications:
– Master’s degree in Statistics, Operations Research, Computer Science, Economics, or related fields.
– Strong quantitative and programming skills. Python programming and knowledge of modern methods (pytorch, polars, spark, arrow, duckdb) is a strong plus.
– Excellent written and verbal communication abilities.
– Demonstrated interest in banking and financial applications.

Application Process:
1. Submit your CV and academic transcripts by December 2nd, 2024 to cbravoro@uwo.ca. Please mention your GRE scores and TOEFL/IELTS or similar tests, if available.
2. Selected candidates will be invited for interviews that same week.
3. Successful interviewees will be invited to submit a formal application to Western University.
4. Final acceptance will be determined by the Graduate Affairs Committee.

To apply or for more information, please contact me at cbravoro@uwo.ca.

The Banking Analytics Lab values diversity and encourages applications from all qualified individuals, including women, members of visible minorities, Indigenous peoples, and persons with disabilities. This position is available to anyone, worldwide.

At least one of these positions will be cosupervised by Dr. María Óskarsdóttir at Southampton University. The job post is below:

https://lnkd.in/gZ92cxaZ

Our latest preprint is out! Check it out on ArXiV.

Imagine making important decisions—like adjusting credit limits or recommending treatments—where there are multiple options instead of only “yes or no”. In our latest research, we asked whether traditional methods of estimating causal effects are enough for these complex situations. But what exactly is a causal effect? Simply put, a causal effect measures how one action directly impacts an outcome—like how increasing a credit limit might influence a customer’s spending behavior. Typically, these causal estimates help us make the “best” decision, but we found they don’t always tell the whole story.

Our study dives into what’s called a “multitreatment scenario”, where the range of possible actions isn’t limited to just one choice. For example, adjusting a credit limit could involve not only deciding whether to change it, but also determining by how much. In our case, the outcome we aimed to optimize was continuous—for instance the expected profit from adjusting credit card limits, and also the outcome was observable both before and after treatment.

We explored a new method that combines causal effect estimation with additional criteria like uncertainty measures and a “prediction condition”. To truly enhance decision quality, we went beyond the traditional methods and added a fresh perspective. We brought in a measure called Conditional Value at Risk (CVaR), which allowed us to understand potential worst-case scenarios, not just the average outcome. Think of CVaR as a safety net that ensures we don’t overlook possible downsides while aiming for the best result.

We also introduced a “predictive condition” to help ensure our recommendations would actually be beneficial. Essentially, the predictive condition means we only recommend an action if we can confidently predict that the outcome will improve compared to where it started. In the context of adjusting credit limits, this meant that any suggested change had to show a clear increase in expected profit compared to the one corresponded to the original limit. This safeguard adds another layer of confidence, making sure that each decision made was a true upgrade rather than just a guess.

When applied to financial decision-making—adjusting credit card limits for consumers—we found that this combined approach outperformed traditional models. It wasn’t just about predicting the treatment effect, but also understanding how confident we could be about that effect while balancing it against other possible outcomes.

What makes this important is its broader applicability. Whether it’s deciding a financial policy, optimizing healthcare treatment doses, or managing resources, this refined approach could change how we navigate complex decisions. By blending causal analysis with a richer understanding of uncertainty, we can make recommendations that are not just statistically sound, but also practically more effective and personalized.

Our study is a step towards smarter decision-making in areas where “one size fits all” just doesn’t cut it. This work helps ensure that recommendations, whether in finance, healthcare, or any other field, are optimized not just for an average effect but for the best possible outcome across a range of real-world scenarios.

You can also listen to a machine-generated podcast of the preprint below.

Older Posts

Blog

WE ARE HIRING: Four Fully Funded PhD Positions in Banking Analytics / Climate Risk!

A Multimodal Approach to SME Credit Scoring Integrating Transaction and Ownership Networks – New Preprint Available!

CBC Weekend Business Panel – October 4th, 2025

Deep Learning in Banking – Preorder our new book!

Our Latest Paper: Multi-modal deep learning for credit rating prediction using text and numerical data streams

CBC News’ Weekend Business Panel – Trump’s Tariffs, Part 2, and the HST Holiday End

CBC News’ Weekend Business Panel – Trump’s Tariffs and TikTok

CBC News’ Weekend Business Panel – The Canada Post strike, the GST Holiday and the latest inflation numbers

Ph.D. Recruiting 2024: Deadline December 2nd!

New Preprint! Are Causal Effect Estimations Enough for Optimal Recommendations Under Multitreatment Scenarios?

The Banking Analytics Lab