Polymorphisms in nucleotide excision repair (NER) genes may cause variations in DNA repair capacity and increase susceptibility to bladder cancer through complex gene-gene and gene-smoking interactions. We applied two data mining approaches to explore high-order gene-gene and gene-environment interactions among 13 polymorphisms in nine major NER genes in 696 bladder cancer patients and 629 controls. Individually, only the XPD D312N variant genotypes exhibited a slightly increased risk for bladder cancer. In classification and regression tree analysis, we observed gene-gene interactions among CCNH V270A, ERCC6 M1097V and RAD23B A249V in ever smokers: smokers with the variant alleles at these three loci had an almost 30-fold increased risk of bladder cancer [odds ratio (OR): 29.6, 95% confidence interval (CI): 9.3-93.7]. When evaluating combined effect of above four single nucleotide polymorphisms, we found a significant gene dosage effect for increased bladder cancer risk with increasing numbers of unfavorable genotypes. Compared with individuals with less than 2 unfavorable genotypes, those with 2 unfavorable genotypes and more than 2 unfavorable genotypes exhibited increased bladder cancer risk with ORs of 1.14 (95% CI: 0.87-1.51) and 2.15 (95% CI: 1.56-2.97), respectively (P < 0.001). The risks were more evident in ever smokers with ORs of 1.43 (95% CI: 1.02-2.01) and 3.40 (95% CI: 2.24-5.15), respectively (P < 0.001). In multifactor dimensionality reduction (MDR) analysis, the five-factor model including smoking, CCNH V270A, ERCC6 M1097V, RAD23B A249V and XPD D312N had the best ability to predict bladder cancer risk. The contributions of these polymorphisms may jointly affect bladder cancer risk through gene-gene and gene-smoking interactions.