Bridges play a central role in transportation systems, and keeping them in good condition is essential for safety and reliable mobility. Traditional inspection methods, although still widely used, depend heavily on visual judgment and can be slow and inconsistent. This paper presents a databased method for evaluating the condition of key bridge components such as the deck, superstructure, and substructure. About 1,650 bridges were examined using information on their structure, geometry, and traffic to estimate the condition of each main component. Two main models were tested: the Support Vector Machine (SVM) and the Random Forest Classifier (RFC). In addition, Support Vector Regression (SVR) was applied to predict condition values in a continuous form. To deal with unbalanced data, both simple and weighted versions of the models were trained. Weighting slightly reduced overall accuracy, but it made the models better at identifying bridges that were in poorer condition. The evaluation of the models was done through performance metrics such as accuracy, precision, recall and F1-score. Feature importance was also analyzed to better understand which parameters had the greatest influence. Overall, the findings show that data-driven evaluation can improve bridge inspection practices, particularly in Morocco, where the use of only visual inspection is common.