John R. J. Thompson
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Research Interests
John R. J. Thompson has a total of 26 co-authors. Here are the top co-authors by number of shared publications:
- Jesse S. Ghashti
Department of Computer Science, Mathematics, Physics and Statistics, University of British Columbia, Kelowna, Canada
3 shared publications - Willard John Braun
University of British Columbia
3 shared publications - Chuck Grace
Finance Ivey Business School London Ontario Canada
2 shared publications - Geoffrey Geotz
Canadian Forest Service, National Resources of Canada, Ottawa, ON K1A 0E4, Canada
2 shared publications - Ladan Tazik
Department of Computer Science, Physics, Mathematics and Statistics, University of British Columbia, Kelowna, BC V1V 1V7, Canada
2 shared publications - Longlong Feng
Financial Wellness Lab Western University London Ontario Canada
2 shared publications
Latest Publications
- Open Access
- 4 May 2026
- Published by Elsevier in Computer-Aided Civil and Infrastructure Engineering
- Vol. 49
Optimal network topology and alignment design for minimizing construction costs of wind farm access roads
- 4 May 2026
- Published by Elsevier in Computer-Aided Civil and Infrastructure Engineering
- Vol. 49
Open AccessOptimal network topology and alignment design for minimizing construction costs of wind farm access roads
- 17 November 2025
- Published by With Intelligence LLC in The Journal of Portfolio Management
- Vol. 52 (2) , 150-183
Clustering and Similarity Learning in Financial Markets: A Tutorial for the Practitioners
Clustering and similarity learning are increasingly indispensable for structuring heterogeneous financial data and supporting real-world decision-making. Traditional heuristics such as industry codes, static style boxes, or return correlations offer only coarse and rigid notions of peer groups. Recent advances in metric learning, graph methods, and large language models now make it possible to build adaptive neighborhoods of securities, funds, companies, and investors that align more closely with actual risk, liquidity, and thematic exposures. This tutorial synthesizes these methodological developments and demonstrates their use across major asset classes. Case studies show how supervised proximities improve bond substitution, how fund similarity systems reconcile category reproducibility with outlier detection, how multimodal pipelines refine company comparables for valuation and strategy, and how investor clustering enhances personalization and “know your client” (KYC) analytics. We emphasize modeling choices that make clustering and similarity auditable and robust under regime shifts. We also outline their evaluation protocols such as neighborhood stability, substitution fidelity, and segment utility, and so on, which align with investment, compliance, and fiduciary objectives. Overall, the central message for practitioners is pragmatic: Similarity systems have moved beyond experimental prototypes and now stand as deployable techniques within real investment workflows.
- 17 November 2025
- Published by With Intelligence LLC in The Journal of Portfolio Management
- Vol. 52 (2) , 150-183
Clustering and Similarity Learning in Financial Markets: A Tutorial for the Practitioners
Clustering and similarity learning are increasingly indispensable for structuring heterogeneous financial data and supporting real-world decision-making. Traditional heuristics such as industry codes, static style boxes, or return correlations offer only coarse and rigid notions of peer groups. Recent advances in metric learning, graph methods, and large language models now make it possible to build adaptive neighborhoods of securities, funds, companies, and investors that align more closely with actual risk, liquidity, and thematic exposures. This tutorial synthesizes these methodological developments and demonstrates their use across major asset classes. Case studies show how supervised proximities improve bond substitution, how fund similarity systems reconcile category reproducibility with outlier detection, how multimodal pipelines refine company comparables for valuation and strategy, and how investor clustering enhances personalization and “know your client” (KYC) analytics. We emphasize modeling choices that make clustering and similarity auditable and robust under regime shifts. We also outline their evaluation protocols such as neighborhood stability, substitution fidelity, and segment utility, and so on, which align with investment, compliance, and fiduciary objectives. Overall, the central message for practitioners is pragmatic: Similarity systems have moved beyond experimental prototypes and now stand as deployable techniques within real investment workflows.
- Open Access
Stochastic Behaviour of Directional Fire Spread: A Segmentation-Based Analysis of Experimental Burns
Understanding the dynamics of fire propagation is essential in improving predictive models and developing effective fire management strategies. This study applies computer vision techniques to complement traditional fire behaviour modelling. We employ the Segment Anything Model to achieve the accurate segmentation of experimental fire videos, enabling the frame-by-frame segmentation of fire perimeters, quantification of the rate of spread in multiple directions, and explicit analysis of slope effects. Our laboratory experiments reveal that the ROS increases exponentially with slope, but with coefficients differing from those prescribed in the Canadian Fire Behaviour Prediction System, reflecting differences in field conditions. Complementary field data from prescribed burns in coniferous fuels (C-7) further demonstrate that slope effects vary under operational conditions, suggesting field-dependent dynamics not fully captured by existing deterministic models. Our experiments show that, even under controlled laboratory conditions, substantial variability in spread rate is observed, underscoring the inherent stochasticity of fire spread. Together, these findings highlight the value of vision-based perimeter extraction in generating precise spread data and reinforce the need for probabilistic modelling approaches that explicitly account for uncertainty and emergent dynamics in fire behaviour.
Open AccessStochastic Behaviour of Directional Fire Spread: A Segmentation-Based Analysis of Experimental Burns
Understanding the dynamics of fire propagation is essential in improving predictive models and developing effective fire management strategies. This study applies computer vision techniques to complement traditional fire behaviour modelling. We employ the Segment Anything Model to achieve the accurate segmentation of experimental fire videos, enabling the frame-by-frame segmentation of fire perimeters, quantification of the rate of spread in multiple directions, and explicit analysis of slope effects. Our laboratory experiments reveal that the ROS increases exponentially with slope, but with coefficients differing from those prescribed in the Canadian Fire Behaviour Prediction System, reflecting differences in field conditions. Complementary field data from prescribed burns in coniferous fuels (C-7) further demonstrate that slope effects vary under operational conditions, suggesting field-dependent dynamics not fully captured by existing deterministic models. Our experiments show that, even under controlled laboratory conditions, substantial variability in spread rate is observed, underscoring the inherent stochasticity of fire spread. Together, these findings highlight the value of vision-based perimeter extraction in generating precise spread data and reinforce the need for probabilistic modelling approaches that explicitly account for uncertainty and emergent dynamics in fire behaviour.
- Open Access
- 14 August 2025
- Published by MDPI AG in Preprints.org
Stochastic Behaviour of Directional Fire Spread: A Segmentation-Based Analysis of Experimental Burns
Understanding the dynamics of fire propagation is essential for improving predictive models and developing effective fire management strategies. This study examines the variability in temporal and directional rates of spread (ROS) under controlled environmental conditions and investigates the influence of terrain slope using experimental fire videos. To enable precise, frame-by-frame tracking of fire perimeters, we employed the Segment Anything Model (SAM) for semantic segmentation and object tracking, allowing us to quantify fire spread. Our study highlight that ROS exhibited substantial variability across and within videos that underscore the stochastic nature of fire behavior and raise concerns about the limitations of deterministic fire spread models. Analysis of the slope spread factor revealed discrepancies between model predictions and observed fire behavior. Estimated slope parameters deviated from values reported in existing literature, suggesting that fire dynamics are highly context-dependent and sensitive to local conditions. Our work highlights the need for probabilistic modeling approaches that explicitly account for inherent uncertainty and emergent dynamics in fire spread. Future research should focus on integrating directional drivers, refining slope-response formulations, and incorporating stochastic processes such as spotting. These improvements are essential for building more robust and generalizable fire behavior models capable of supporting operational forecasting and management decisions.
- 14 August 2025
- Published by MDPI AG in Preprints.org
Open AccessStochastic Behaviour of Directional Fire Spread: A Segmentation-Based Analysis of Experimental Burns
Understanding the dynamics of fire propagation is essential for improving predictive models and developing effective fire management strategies. This study examines the variability in temporal and directional rates of spread (ROS) under controlled environmental conditions and investigates the influence of terrain slope using experimental fire videos. To enable precise, frame-by-frame tracking of fire perimeters, we employed the Segment Anything Model (SAM) for semantic segmentation and object tracking, allowing us to quantify fire spread. Our study highlight that ROS exhibited substantial variability across and within videos that underscore the stochastic nature of fire behavior and raise concerns about the limitations of deterministic fire spread models. Analysis of the slope spread factor revealed discrepancies between model predictions and observed fire behavior. Estimated slope parameters deviated from values reported in existing literature, suggesting that fire dynamics are highly context-dependent and sensitive to local conditions. Our work highlights the need for probabilistic modeling approaches that explicitly account for inherent uncertainty and emergent dynamics in fire spread. Future research should focus on integrating directional drivers, refining slope-response formulations, and incorporating stochastic processes such as spotting. These improvements are essential for building more robust and generalizable fire behavior models capable of supporting operational forecasting and management decisions.
- 31 May 2025
- Published by Springer Nature in Advances in Data Analysis and Classification
- p. 1-37
Mixed-type kernel-based quantification of similarity for clustering
- 31 May 2025
- Published by Springer Nature in Advances in Data Analysis and Classification
- p. 1-37
Mixed-type kernel-based quantification of similarity for clustering
- Open Access
- 14 November 2024
- Published by Association for Computing Machinery (ACM)
- p. 186-194
Functional Mixed-type Clustering of Investors' Daily Returns During a Market Shock Change-point and Recovery
We analyze a dataset of daily portfolio market positions spanning July 17th 2019 to February 19th 2021 for many client accounts of a financial dealer. For each client, we know demographic information (age, gender, marital status, income, retirement status, investment knowledge), desired risk preferences, and their market positions for July 17th 2019 to February 19th 2021. The daily unrealized returns of those portfolios are analyzed using a functional principal component approach to discover underlying variations in observed client returns and if those differences in portfolios resulted in different outcomes before, during, and after the short-duration 2020 market crash. The distinguishing factors between portfolio returns are, in the order of importance, performances during the crash, before the crash, overall throughout, during the recovery, and mitigating losses at the sacrifice of recovery. We conduct mixed-type categorical analysis by mixing 15 functional principal components with categorical data using k-prototypes clustering. Clients are grouped into six categories: two small groups with strong performances corresponding to the volatility of their positions and four large groups, all with smaller volatilities and a range of overall performances. We find that clients tended to overshoot their risk appetites in clusters: high-risk appetites tended toward over-risked; low-risk tended toward under-risk. Overall, performance was connected to increased risk throughout the market crash—if the clients did not take as much risk, they did not experience as strong a recovery.
- 14 November 2024
- Published by Association for Computing Machinery (ACM)
- p. 186-194
Open AccessFunctional Mixed-type Clustering of Investors' Daily Returns During a Market Shock Change-point and Recovery
We analyze a dataset of daily portfolio market positions spanning July 17th 2019 to February 19th 2021 for many client accounts of a financial dealer. For each client, we know demographic information (age, gender, marital status, income, retirement status, investment knowledge), desired risk preferences, and their market positions for July 17th 2019 to February 19th 2021. The daily unrealized returns of those portfolios are analyzed using a functional principal component approach to discover underlying variations in observed client returns and if those differences in portfolios resulted in different outcomes before, during, and after the short-duration 2020 market crash. The distinguishing factors between portfolio returns are, in the order of importance, performances during the crash, before the crash, overall throughout, during the recovery, and mitigating losses at the sacrifice of recovery. We conduct mixed-type categorical analysis by mixing 15 functional principal components with categorical data using k-prototypes clustering. Clients are grouped into six categories: two small groups with strong performances corresponding to the volatility of their positions and four large groups, all with smaller volatilities and a range of overall performances. We find that clients tended to overshoot their risk appetites in clusters: high-risk appetites tended toward over-risked; low-risk tended toward under-risk. Overall, performance was connected to increased risk throughout the market crash—if the clients did not take as much risk, they did not experience as strong a recovery.
- 7 October 2024
- Published by Springer Nature in Journal of Classification
- Vol. 42 (2) , 311-334
Mixed-Type Distance Shrinkage and Selection for Clustering via Kernel Metric Learning
- 7 October 2024
- Published by Springer Nature in Journal of Classification
- Vol. 42 (2) , 311-334
Mixed-Type Distance Shrinkage and Selection for Clustering via Kernel Metric Learning
- 27 August 2024
- Published by The R Foundation
kdml: Kernel Distance Metric Learning for Mixed-Type Data
- 27 August 2024
- Published by The R Foundation
kdml: Kernel Distance Metric Learning for Mixed-Type Data
- Open Access
- 3 July 2024
- Published by The R Foundation
nonsmooth: Nonparametric Methods for Smoothing Nonsmooth Data
- 3 July 2024
- Published by The R Foundation
Open Accessnonsmooth: Nonparametric Methods for Smoothing Nonsmooth Data
- 21 May 2024
- Published by Taylor & Francis in Journal of Applied Statistics
- Vol. 51 (16) , 3431-3455
Iterative smoothing for change-point regression function estimation
Understanding wildfire spread in Canada is critical to promoting forest health and protecting human life and infrastructure. Quantifying fire spread from noisy images, where change-point boundaries separate regions of fire, is critical to accurately estimating fire spread rates. The challenge lies in denoising the fire images and accurately identifying highly non-linear fire lines without smoothing over boundaries. In this paper, we develop an iterative smoothing algorithm for change-point data that utilizes oversmoothed estimates of the underlying data generating process to inform re-smoothing. We demonstrate its effectiveness on simulated one- and two-dimensional change-point data, and robustness to response outliers. Then, we apply the methodology to fire spread images from laboratory micro-fire experiments and show that the regions fuel, burning and burnt-out are smoothed while boundaries are preserved.
- 21 May 2024
- Published by Taylor & Francis in Journal of Applied Statistics
- Vol. 51 (16) , 3431-3455
Iterative smoothing for change-point regression function estimation
Understanding wildfire spread in Canada is critical to promoting forest health and protecting human life and infrastructure. Quantifying fire spread from noisy images, where change-point boundaries separate regions of fire, is critical to accurately estimating fire spread rates. The challenge lies in denoising the fire images and accurately identifying highly non-linear fire lines without smoothing over boundaries. In this paper, we develop an iterative smoothing algorithm for change-point data that utilizes oversmoothed estimates of the underlying data generating process to inform re-smoothing. We demonstrate its effectiveness on simulated one- and two-dimensional change-point data, and robustness to response outliers. Then, we apply the methodology to fire spread images from laboratory micro-fire experiments and show that the regions fuel, burning and burnt-out are smoothed while boundaries are preserved.