The distribution of grain sizes, or the distribution of particle sizes within a reservoir, is an important parameter for the petroleum industry. This can help determine areas that have the best production potential, including optimizing the well completion and production strategies. Other purposes for grain size distribution are formation evaluation and production optimization, which also contributes to further optimizing production and minimizing reservoir damage. This paper presents a study that demonstrates the feasibility of using machine learning to predict grain size distribution from well logs using a Random Forest algorithm. Training of the algorithm was performed using well logs that had a high correlation with the weighted average grain size calculated per depth interval for the whole well intervals. The model achieved an R² score of 0.89, which is a promising starting point for future improvements. This paper also highlights the use of statistical methods to improve the accuracy of machine learning algorithms. One of these methods calculates the average value of the adjacent depth interval and reduces the impact of undetected outliers.