Explainable fuzzy logic for sustainable housing investment risk assessment: evidence from Türkiye

Purpose The paper aims to examine whether Mamdani-type fuzzy logic inference can serve as an explainable artificial intelligence (XAI) framework for housing investment risk assessment in real estate valuation and finance. Using building age, floor area and unit price as inputs, the model produces interpretable risk scores for seven major Turkish cities (Istanbul, Ankara, Izmir, Bursa, Antalya, Adana and Konya) that can inform discussions of housing market sustainability [sustainable development goal (SDG) 11] and affordability (SDG 10); the model does not directly measure income, exploitation or inequality outcomes. Design/methodology/approach Secondary data comprising 66,725 housing listings from September 2023 were used. Building age, gross square meters and unit price (TL/m²) were defined as input variables, while investment risk score was the output variable. The model, which includes 25 fuzzy IF–THEN rules, was adjusted based on specific city percentile thresholds (P25, P50, P75), allowing the model to account for differences among local housing markets. Findings Clear risk differentiation was detected across cities, consistent with underlying spatial inequalities. Izmir (average risk score: 57.0 on a 0–100 scale; high-risk rate: 41.3%) and Istanbul (high-risk rate: 35.6%) show the most pronounced speculative price dynamics, a pattern with potential implications for affordable housing access under SDG 11. Konya (average risk score: 40.4; high-risk rate: 9.6%) shows a more balanced market profile. Building age is the strongest correlate of risk in most cities studied, linking physical housing stock deterioration to financial risk; the exception is Antalya, where unit price shows the strongest association with risk (r = 0.672), suggesting demand-driven rather than age-driven pressure in that market. These patterns carry implications for urban regeneration and energy-efficient building policy. Originality/value This research combines fuzzy logic risk modeling with existing literature on XAI in real estate valuation. The clear 25-rule inference system offers a useful method for recognizing risks in the speculative housing market and may also support valuation education and policy development in line with the SDGs. The multi-city calibration design (n = 66,725) and city-specific membership functions provide an observable and interpretable framework for housing investment risk assessment across urban real estate markets.