In this paper, two models ANN and ANN + PSO were proposed to predict the mechanical properties of Magnesium (Mg) alloys, and the results were compared. Now-a-days, machine learning algorithms are used in various engineering applications in prediction of parameters and properties of materials, machining, and manufacturing. Twenty samples of as-cast Mg alloys with varying weight percentages of its constituents—Aluminum (Al), Zinc (Zn), Manganese (Mn) and Calcium (Ca)—were taken along with their yield strength (YS), ultimate tensile strength (UTS), and ductility (%) to build an artificial neural network (ANN) model for prediction of mechanical properties of Mg-Al-Zn-Ca alloys which play a vital role in making biodegradable implants for orthopedic applications. The weight percentages of Mg, Al, Zn, Mn, and Ca were taken as input data, and YS, UTS, and ductility were considered as target data in training the ANN model. The hidden layer with 2–10 neurons was considered to observe the performance of the model in predicting these properties. Levenberg–Marquart (LM) algorithm was used to train the input data, and transigma function was used as a learning function. The mean square error (MSE) and correlation coefficient (R2) were used to validate the optimality of the results. The weights of the neurons were optimized using particle swam optimization (PSO) to enrich the predicted results from ANN. The results obtained from ANN and ANN + PSO model show that the later performs better in prediction of mechanical properties of Mg-Al-Zn-Ca alloys.