In almost the last two decades, commercial Bridge Management System (BMS) packages have been remarkably developed. However, inconsistency between BMS inputs and bridge agencies' existing data is an obstacle to implement and to operate a BMS software application. A large number of bridge datasets for a BMS database is an essential requirement to analyze a bridge network. Among many requirements, historical structural datasets are vital to compute the prioritization of bridge stock for maintenance and repair activities and are mostly unavailable for bridges of more than 20 years in age. This study focuses on the abovementioned difficulty to overcome the lacking historical data problem faced by bridge agencies to effectively use BMS applications. This paper proposes an artificial neural network (ANN) technique to predict missing components of time-series datasets to estimate historical bridge element condition ratings. Although this study only estimates historical condition ratings, the proposed concept can be used to compute other historical dataset requirements in the BMS database and hence improving the reliability of various BMS analysis modules.