In an industrial Catalytic Reforming Unit Platformer Heater, tube wall temperature is one of the most important variables that dictate the overall efficiency of the heater. Due to the small margin of allowable error of the tube wall temperature, high value of thermocouple tolerance will jeopardize the safety of the Catalytic Reforming Unit Platformer Heater.To overcome this problem, inferential approach is proposed to predict the tube wall temperature.Original data from the plant information system (PI) which consist of 15 variables are filtered using regression analysis in order to obtain the most correlated variables that affect the behavior of the tube wall temperature. Data set that consist of six variables that are strongly correlated are used to develop the prediction model using Neural Network and Projection to Latent Structure (PLS) methods. The results show that Cascade-forward Backpropagation network gives the best model prediction for the tube wall temperature with RMSE of 1.35 and CDC of 83.62%. Predictions of tube wall temperature can be the ‘element’ of the furnace optimization, since the tube wall temperature can be predicted with the lower tolerance, combustion air in excess that act as a cooling medium can be reduced. Reducing the excess air will also reduce the fuel gas consumption and thus increase the platformer heater efficiency. This predictions also may prevent the overshoot temperature of the tube wall and avoiding the possibilities of tube sagging and/or ruptures which is to be the main priority for safety.