The invention discloses a combination forecast method for urban road traffic conditions. According to the combination forecast method, a maximum-minimum method is used for estimating weekly similarity of traffic flows and average speed of a current traffic condition sequence and a traffic condition sequence of the same time on a Sunday in the same week; if the weekly similarity is in an acceptable range, the weekly similarity is used for reckoning a traffic flow and average speed of a next moment; if the weekly similarity is not in the acceptable range, a fractal theory-based prediction algorithm is used for calculating a traffic flow and average speed of predicted time; finally a relative evaluation method is used for estimating a traffic jam index of a current road segment at the predicted time. A fractal dimension algorithm employed in the combination forecast method is based on a structure function, traffic conditions at different time intervals based the predicted time are considered, and restoration is performed via use of a special weighting coefficient. The combination forecast method is advantaged by large computation amount elasticity, high prediction precision and comprehensive output traffic condition data.