Laboratory measurements are the most reliable methods for PVT analysis of reservoir fluid properties. Faster and cheaper methods are important for real time decision making and mathematical correlations are used to predict the properties. Most of the existing correlations do not give accurate estimations due to either simplifying assumptions or inappropriate technique. In this work, a new correlation for estimating oil formation volume factor (βo) using a Group Method of Data Handling (GMDH) technique was developed. It is a family of inductive algorithms which executes computer-based mathematical modeling of multi-parametric data sets. In developing the correlation, 93 PVT data sets from Malaysian crude oil analyses were used. The data consist of solution gas-oil ratio (Rs), βo, oil viscosity (μ,o), oil gravity (γo), gas gravity (γg) and temperature (T). Some of these parameters were set as inputs while the properties to be predicted were set as output. The correlation was developed using the estimated outcome from the technique. Subsequently, 39 data points from different reservoir fluids were used to validate the correlation. Analyses of results show that the GMDH correlation is more accurate than the existing correlations with the Average Absolute Relative Error (AARE) of 0.976, and with maximum and minimum average relative errors (max. ARE and min. ARE) of 3.989 and 0.001 respectively. Copyright © (2014) by the Society of Petroleum Engineers All rights reserved.