Using Supervised Learning to Improve Monte Carlo Integral Estimation