There is an urgent need to forecast real estate unit prices because the average price of residential real estate is always fluctuating. This paper provides a real estate price prediction model based on supervised regression deep learning with 3 hidden layers, a Relu activation function, 100 neurons, and a Root Mean Square Propagation optimizer (RMS Prop). The model was developed using actual data collected from 28 Egyptian cities between 2014 and 2022. The model can forecast the price of a real estate unit based on 27 different variables. The model is created in two stages: adjusting the parameters to obtain the best ones using a sensitivity k-fold technique, then optimizing the result. 85 percent of the real estate unit data gathered was used in training and developing the model, while the other 15 percent was used in validating and testing. By using a dropout regularization technique of 0.60 on the model layers, the final developed model had a maximum error of 10.58%. After validation, the model had a maximum error of about 9.50%. A graphical user interface (GUI) tool is developed to make use of the final predictive model, which is very simple for real estate developers and decision-makers to use.