This work proposes an improved quantitative prediction model for corrosion under insulation (CUI) for oil and gas piping in equatorial climate zone, using real-world data and integrated with experimental work using modified two stages of Artificial Neural Network. Investigation into the effect of single type of data source and different type of fitting line analysis on the final CUI model are also discussed with the goal for Risk Based Inspection (RBI) to be more effective. Results from the CUI prediction model exhibits high R2 of 0.9919 and Root Mean Squared Error (RMSE) of 0.0087. © 2018 Author(s).