Crude oil is made of different compositions and materials, namely hydrocarbons, oxygen, nitrogen, sulfur, and various metals. Deposition of heavy components can lead to a restriction of flow paths and decrease the oil production rate during production processes. Deposition of wax is a crucial issue that appears in the transfer equipment and surface facilities and decreases the inner diameter of pipelines. Wax deposition leads to loss of effective diameter and an increase in roughness, which will increase the pressure drop in the oil reservoir. Deposition of wax particles appears in low temperatures which can lead to a blockage of pipelines. When the oil's temperature is less than the wax appearance temperature (WAT), which is called the cold flow regime, oil composition changes because wax crystals precipitate. This research generally focuses on developing accurate intelligent models to predict wax deposition using the pour point temperature and °API of the crude oil sample as input variables. Various smart networks, including multilayer perceptron (MLP), radial basis function (RBF), cascade forward neural network (CFNN), and generalized regression neural network (GRNN) are utilized for this purpose. Four optimization algorithms, namely Broyden–Fletcher–Goldfarb–Shanno quasi-Newton backpropagation (BFGS), Levenberg-Marquardt (LM), Scaled Conjugate Gradient (SCG), and Bayesian Regularization (BR) are used for training the models. The results showed that the developed GRNN smart model could obtain a superior performance providing predictions for the amount of wax deposited with a root mean squared error (RMSE) of 0.845, an average absolute percent relative error (AAPRE) of 10.36%, and a coefficient of determination (R2) of 0.975. The trend analysis indicated that increasing pour point temperature leads to a wax deposition increase. Oppositely, increasing oil °API gravity reduces wax deposition. Lately, outlier detection was carried out by applying the Leverage approach to identify the outlier data points in a data bank. It detected only 4 data points (out of 173) as outliers.