Abstarct: In this work, a wavelet neural network (WNN) has been used to predict density of ketones over wide range of temperature and pressure. At first, WNN has been constructed using temperature and pressure variants 4 descxriptors baxsed on group contribution method namely, the number of methyl, methylene, methyne, and carbonyl functional groups. Then, a computer program was written MATLAB 6.1 (Math Works) to use wavelet neural network. After training of networks, The WNN parameters consist of the number of neurons in the hidden laxyer, the learning rate, the momentum and the numbers of iterations have been optimized simultaneously for each of ketones. The capability of the optimized models has been evaluated by plotting experimental values of density versus the predicted values for the prediction and validation sets. The obtained result with an average percent deviation error for density prediction lower than 1.3% reveals the capability of the WNN model. To achieve a WNN model for density prediction of all ketones, this approach was successfully applied for simultaneous prediction of 2-butanone and 2-pentanone. The result for density prediction show that an average percent deviation error lowers than 0.6%. After this, a WNN model with 6 descxriptors has been constructed and optimized for density prediction of all ketones. The average percent deviation error and its maximum value for density prediction were found to be lower than 1.05% and 1.95%, respectively. The excellent prediction obtained with a correlation coefficient 0.99 reveals the capability of the WNN models.