Mining Air Pollution Patterns in Monterrey: A Comparative Apriori and FP-Growth Approach

Air pollution in the Monterrey Metropolitan Area, Mexico, is a complex phenomenon driven by interactions among anthropogenic emissions, meteorological conditions, and regional topography. This work aimed to identify co-occurrence patterns between criteria pollutants (NO₂, SO₂, and CO) and meteorological variables (temperature, relative humidity, barometric pressure, and wind speed) by mining association rules with the Apriori and FP-Growth algorithms. Hourly records from 2016–2024 at nine stations of the Nuevo León State (Mexico) atmospheric monitoring system were used. Data were organized by station and time slot and discretized using the 50th percentile as the threshold to classify records as high or low. The comparative analysis showed that both algorithms converged to the same set of rules, but FP-Growth significantly optimized computational efficiency by eliminating the need to generate intermediate candidates. The rules were evaluated in two experiments that varied the minimum support and confidence thresholds, and the top 30 rules with lift values greater than 1 were selected. At 10% support and 90% confidence, critical patterns were identified that associate simultaneous high levels of NO₂ and CO with atmospheric stability scenarios characterized by low wind speeds, low temperatures, and high relative humidity. On the other hand, when the thresholds were relaxed to 30% support and 70% confidence, high-frequency meteorological rules linked to regional climate dynamics predominated. In conclusion, association rule mining proved to be an effective and complementary tool to traditional statistics for characterizing recurrent atmospheric interactions, providing valuable information for co-designing early warning systems and urban air quality management.