The problem is more consequential than it might first appear. Most outlier detection algorithms were designed with numerical data in mind. They measure distances, densities, and deviations in spaces where every attribute is a number, and where concepts like proximity are mathematically straightforward. But real datasets are rarely so cooperative. Medical records contain diagnoses and blood types. Customer databases contain product categories, subscription tiers, and regions. Fraud datasets mix transaction amounts with merchant categories and card types. In these settings, the numerical machinery of classical anomaly detection either breaks down or silently distorts the data, and practitioners have been left to improvise without solid evidence about which improvisation works.

Ferreira and Cordeiro frame the field’s improvisations as three competing strategies. The first is to apply algorithms that can process categorical data directly, respecting the discrete nature of attributes such as color, brand, or diagnosis. The second is to convert categorical attributes into numerical ones before detection, transforming the data into a form that standard numerical detectors can consume. The third is the bluntest instrument: simply remove the categorical attributes and run detection on the numerical columns alone. Each strategy has intuitive appeal and intuitive drawbacks, but until now there has been no rigorous, large-scale comparison of how they actually perform across diverse datasets and detectors.

To settle the question, the researchers mounted an unusually comprehensive experimental campaign. They assembled 47 datasets and ran 14 different outlier detection algorithms across them, evaluating all three strategic approaches under controlled conditions. This scale matters. Many published comparisons in machine learning rest on a handful of benchmark datasets, which makes it dangerously easy to overfit conclusions to particular data quirks. By sweeping across dozens of datasets with varied characteristics, the study could distinguish between strategies that win consistently and strategies that win only under specific conditions, a distinction that turns out to be central to the paper’s findings.

The headline result is that the first strategy, applying algorithms that handle categorical data natively, is usually the preferred choice. And within that strategy, one detector stood out from the pack: CBRW, an algorithm that approaches categorical outliers through a weighted-graph perspective on attribute values. When datasets contain categorical attributes, the evidence indicates that reaching for a purpose-built categorical detector, particularly CBRW, is the most reliable default. This finding carries real practical weight, because it suggests that the common habit of forcing categorical data through numerical pipelines may often be an unnecessary detour that costs detection quality.

But the study resists the temptation of a one-size-fits-all verdict. The second strategy, converting categorical attributes into numbers before detection, proved superior in certain contexts, and the researchers found that which contexts depends on the characteristics of the data itself. Notably, when this conversion route was taken, detectors such as Isolation Forest, known as iForest, and KNN-outlier achieved better results in those favorable settings. Isolation Forest works by randomly partitioning the data and isolating records that are easy to separate from the rest, while KNN-outlier flags points that sit far from their nearest neighbors. Both are staples of the numerical anomaly detection toolkit, and the study shows they can still earn their keep on mixed data, provided the categorical-to-numerical conversion is done well.

That proviso led the authors to a second, finer-grained comparison: among the many techniques for converting categorical attributes into numerical representations, which one should practitioners choose? Here again the study produced a clear signal. Correspondence Analysis, a classical statistical method that maps categorical variables into a low-dimensional numerical space while preserving the associations between attribute values, often yielded the best detection results among the conversion approaches evaluated. For teams committed to running numerical detectors on mixed data, this finding offers a concrete, evidence-backed recommendation in place of an arbitrary choice of encoding technique.

Perhaps the most forward-looking contribution goes beyond ranking the strategies. Because the best approach depends on data characteristics, the researchers built a predictive model, a meta-level classifier trained on their experimental findings, that examines a new dataset and predicts which of the three strategies is most likely to perform best. The model achieves 80 percent accuracy in identifying the winning strategy. In a field where practitioners often have no principled basis for choosing an approach, an 80 percent reliable recommendation engine represents a meaningful step toward automated, data-driven method selection, analogous to the meta-learning systems that have begun to guide algorithm choice in supervised learning.

The study also delivers a quieter but equally important negative result: the third strategy, discarding categorical attributes altogether, does not emerge as a recommended path. Throwing away information is the kind of shortcut that survives in practice because it is easy, and part of the value of a systematic comparison is that it can retire such shortcuts with evidence. The findings collectively suggest that the information carried by categorical attributes is genuinely useful for anomaly detection, whether consumed directly by a categorical-aware algorithm or preserved through a careful numerical conversion such as Correspondence Analysis.

For the broader machine learning community, the significance of this work lies in its methodological honesty. Rather than crowning a single universal champion, it maps the conditions under which each strategy excels and then builds a bridge from those conditions to practical recommendations for new data. The authors have also committed fully to reproducibility: all code, detailed results, parameter values tested, and datasets used in the study are freely available for download on GitHub, allowing other researchers to scrutinize, extend, or challenge the conclusions. Supported in part by the Brazilian agency CAPES and published open access, the survey arrives as both a practical handbook for anyone deploying anomaly detection on messy real-world data and a benchmark against which future categorical outlier detection methods will now have to be measured. As fraud, cybersecurity, and data-quality applications continue to multiply across industries dominated by categorical information, the question this study answers is no longer an academic curiosity. It is an operational decision worth getting right, and for the first time, practitioners have a large-scale evidence base to guide it.

Subject of Research: Comparative evaluation of outlier detection methods for categorical and mixed data

Article Title: A comparative evaluation of outlier detection in categorical and mixed data

Article References: Ferreira, F. P., & Cordeiro, R. L. F. (2026). A comparative evaluation of outlier detection in categorical and mixed data. Artificial Intelligence Review. https://doi.org/10.1007/s10462-026-11687-3

Image Credits: AI Generated

DOI: 10.1007/s10462-026-11687-3

Keywords: outlier detection, categorical data, mixed data, anomaly detection, CBRW, Isolation Forest, KNN-outlier, Correspondence Analysis, data conversion, machine learning, data mining, benchmark evaluation

The research team, led by geographers at Adamas University in West Bengal, used the Local Climate Zone (LCZ) framework, a classification scheme introduced by urban climatologists Ian Stewart and Tim Oke in 2012. Instead of treating a city as a single hot mass, the LCZ scheme divides it into zones of uniform surface cover, structure, material, and human activity, spanning hundreds of meters to several kilometers. Compact high-rise districts, lightweight slum housing, heavy industry, dense tree cover, and low agricultural plants each form distinct zones with their own thermal signatures. This matters because the urban heat island effect, first noticed by Luke Howard in London in the 1810s, is not uniform: it is the product of street geometry, building height, impervious surfaces, and the slow disappearance of vegetation that shades and cools.

To map these zones, the researchers turned to the free Landsat satellite archive, downloading every available scene from 2010, 2015, 2020, and 2023 with less than ten percent cloud cover. Land surface temperature was retrieved using a single-channel algorithm for Landsat 5 and a split-window method for Landsat 8, both executed in Google Earth Engine. The LCZ classification itself was performed in the open-source SAGA-GIS platform using a supervised Random Forest classifier trained on carefully refined sample polygons drawn from high-resolution Google Earth imagery, following the WUDAPT Level 0 protocol. Landsat scenes were resampled from 30 meters to a 100-meter grid, the scale at which local urban structure is best represented, and majority filtering removed isolated pixels to produce morphologically coherent zones.

The classification proved robust. Kappa coefficients ranged from 0.8367 in 2010 to 0.8462 in 2023, with overall accuracy between 83 and 84 percent in every year. The team went further than most satellite studies by validating the maps on the ground: they captured fisheye photographs of the sky at breast height across six sampled locations, from the planned avenues of Kalyani to the congested lanes of the Kolkata Municipal Corporation area. By converting these images to black and white, separating sky pixels from building, road, and tree pixels, and computing the ratio of open sky to the full circular image, they derived Sky View Factors for each site. The measured values matched the documented morphology of the assigned LCZ classes, confirming that the satellite-derived zones reflect real street-canyon geometry, which governs how much solar radiation reaches the ground and how much longwave radiation escapes at night.

The temporal picture is stark. In 2010, building-type zones covered roughly 40 percent of the metropolitan area while vegetated zones held 53 percent. By 2020 the balance had tipped, with built forms at 52 percent and vegetation down to 41 percent. By 2023, built-up classes dominated at 66.37 percent of the total area, and vegetated classes had collapsed to just 12.45 percent. The city has not merely spread; it has compacted. Open high-rise and open mid-rise forms were converted into compact mid-rise (LCZ 2) and compact low-rise (LCZ 3) fabric, signaling both vertical growth and densification. Mean centers of the zone distributions reveal the direction of change: compact high-rise and mid-rise zones shifted southward toward Sonarpur Rajpur, where scarce land pushed developers toward gated high-rise apartments, while most other built classes expanded northward and northeastward into Barasat, Barrackpore, and Kamarhati, where open land remained. The southward retreat of dense trees, scattered trees, scrub, and low plants marks the loss of green and agricultural land in the north.

Land surface temperatures rose accordingly. The annual mean climbed from about 21.99 degrees Celsius in the late reference period to 23.25 degrees Celsius in 2023, and the Landsat retrievals correlated strongly (r = 0.80) with the MODIS Terra daily temperature product, lending independent credibility to the estimates. When the researchers computed heat island intensity, the difference between each built zone’s mean temperature and that of low plants (LCZ D), the rural reference, three classes stood out: lightweight low-rise (LCZ 7), large low-rise (LCZ 8), and heavy industry (LCZ 10). These zones, home to low-cost slum housing and the jute, textile, and chemical industries along the Hooghly River belt, consistently showed the strongest and most persistent surface heat island signatures across all four study years.

But temperatures on a map do not sweat. To capture the human dimension, the team surveyed 715 adults across 28 municipalities in March to May 2023, using stratified random sampling across all nine building-type LCZs. Every respondent reported at least one hour of daily outdoor sun exposure, divided into four bands from one to three hours up to ten to twelve hours. The questionnaire catalogued 28 heat-induced discomforts, from headaches and heavy sweating to dizziness, excessive thirst, blurred vision, skin rashes, back pain, muscle pain, and even seizures. Geocoding each residence and overlaying the points on the 100-meter LCZ map linked every report to a specific urban morphology. A heat map of relative frequencies showed that ten symptoms were moderate to highly frequent regardless of exposure time, with headache and heavy sweating dominant everywhere, and that the probability of nearly every discomfort rose steadily as hours in the sun increased.

The most technically innovative step was a correspondence analysis linking LCZ classes to reported symptoms for each exposure band, tested with chi-square statistics that were significant at p < 0.001 for all four durations. The pattern changed with time. During one-to-three-hour exposure, the associations were noisy: lightweight low-rise zones and large low-rise zones appeared as outliers, and symptoms clustered loosely without clear zone-specific links. As exposure lengthened, the structure sharpened. Total inertia, a measure of how differentiated the zone-symptom relationships are, rose from 0.215 for short exposure to 0.584 for ten-to-twelve-hour exposure, while the first two dimensions explained up to 76.5 percent of the variance. By the longest exposure band, specific pairings emerged: heavy sweating with compact low-rise zones, heat stroke with compact high-rise, suffocation with compact high-rise and mid-rise, and headache and back pain with open mid-rise and heavy industrial zones. Prolonged heat, in other words, converts generalized misery into a predictable geography of illness.

The implications reach well beyond Kolkata. The findings align with LCZ studies in Hamburg, Nanjing, and central European cities showing that densification raises surface temperatures, but they add something most such studies lack: empirical, field-based evidence of who suffers and how. Because lightweight low-rise settlements and industrial belts combine the highest temperatures with some of the city’s most vulnerable residents, the study exposes a socio-spatial dimension of urban warming, where those least able to afford cooling bear the greatest thermal burden. The authors note that the zones shifting northward and northeastward, including open mid-rise, open low-rise, lightweight low-rise, and industrial classes, deserve priority planning attention, and they point to artificial intelligence, green infrastructure, and reflective surfaces as tools for predictive modeling and mitigation aligned with the Sustainable Development Goals on sustainable cities, climate action, and health.

What makes this research resonate is its refusal to separate climate science from lived experience. A city that loses two-fifths of its vegetation in thirteen years does not just warm statistically; its residents report more headaches, more dizziness, more thirst, and more dangerous symptoms as their hours outdoors lengthen. The lesson for rapidly urbanizing tropical cities is that urban form is health policy. Every compact mid-rise tower raised on a former wetland, every tree line cleared for a road, quietly rewrites the thermal contract between a city and the people who walk its streets, and the new study gives planners a framework, validated from orbit down to the fisheye lens, for reading that contract before it is signed.

Subject of Research: Heat-induced health discomforts associated with changing local climate zones in the Kolkata Metropolitan Area, India

Article Title: Understanding the heat-induced discomforts associated with changing local climate zones in indian cities

Article References: Parial, M., Khorat, P., Mukherjee, K., Dash, P., Karmakar, S., Roy, A. D., Bhadra, T., Saha, A., & Dhali, B. (2026). Understanding the heat-induced discomforts associated with changing local climate zones in indian cities. Discover Cities, 3(1), Article 190. https://doi.org/10.1007/s44327-026-00288-y

Image Credits: AI Generated

DOI: 10.1007/s44327-026-00288-y

Keywords: urban heat island, local climate zones, Kolkata, land surface temperature, remote sensing, sky view factor, urban morphology, heat stress, public health, correspondence analysis, urbanization, West Bengal