In this study, water absorption and thickness swelling values of medium density fiberboard (MDF) were modelled by artificial neural networks (ANN). MDF panels were produced with different rates of paraffin (0.0-control, 0.5, 1 and 1.5 %) at different press temperatures (170 and 190 °C). After conditioning of MDF, water absorption (WA) and thickness swelling (TS) of samples were carried out at specific intervals within 24 hours. Then, the data obtained from these experiment were modelled using ANN. Paraffin addition rate, press temperature and immersion time in water were used as the input parameters, while WA and TS values of MDF were used as the output parameters. After training of ANN, it was found that correlation coefficients (R) were close to 1 for training, validation, test and all data set. Mean absolute percentage error (MAPE) and mean square error (MSE) were determined as 2.94 % and 0.57, respectively, for all data sets. As a result of this study, the use of proposed ANN model may be recommended to predict the water absorption and thickness swelling of panels instead of complex and time-consuming studies such as empirical formulas.; U istraživanju je modelirana upojnost vode i debljinsko bubrenje ploče vlaknatice srednje gustoće (MDF ploče) uz pomoć umjetnih neuronskih mreža (ANN-a). MDF ploče proizvedene su uz dodatak različitih količina parafi na (0,0 – kontrola, 0,5; 1 i 1,5 %) pri različitim temperaturama prešanja (170 i 190 °C). Nakon kondicioniranja MDF ploče, mjerena je upojnost vode (WA) i debljinsko bubrenje (TS) uzoraka u određenim intervalima unutar 24 sata. Zatim su ti podatci modelirani uz pomoć ANN-a. Kao ulazni parametri poslužili su količina parafi na, temperatura prešanja i trajanje namakanja uzoraka u vodi, dok su WA i TS vrijednosti MDF ploče korištene kao izlazni parametri. Nakon provedbe ANN-a utvrđeno je da su koeficijenti korelacije (R) za provedbu, validaciju, ispitivanje i sve skupove podataka blizu 1. Srednja apsolutna pogreška (MAPE) i srednja kvadratna pogreška (MSE) za sve su skupove podataka iznosile 2,94 % i 0,57. Kao rezultat ovog istraživanja može se preporučiti uporaba predloženog ANN modela za predviđanje upojnosti vode i debljinskog bubrenja ploča umjesto složenih i dugotrajnih studija poput empirijskih formula.