Today, different software exists to Predict the Likelihood of Failure (LoF) for the pipes in the water networks. The LoF is used to prioritize the pipes for renovation planning of the water networks. The modern techniques use statistical and/orMachine Learning models to predict the LoF. However, most of this software needs the historical bursts records of the network to predict the LoF. The historical bursts records are unavailable or not well recorded in many water networks, thus, this kind of software could not be used for LoF prediction. In this project, we have attempted to fill this gap and build a data-driven LoF model for water networks lacking their historical bursts records. In the first part of this project, we have created a database of pipes with their historical bursts records from SUEZ water networks in different countries. Then, we have enriched the SUEZ data with open-source data to add supplementary information to each pipe. Afterwards, we have built aMachine Learningmodel (Generic Model) to predict the Lof for networks without the historical bursts records. For a new water network, Generic Model finds similar pipes in the SUEZ database. It trains a classification algorithm on these similar pipes to predict the LoF for the pipes of the new network.