A Fast Parallel Random Forest Algorithm Based on Spark

To improve the computational efficiency and classification accuracy in context of big data, an optimized parallel random forest algorithm is proposed based on Spark computing framework. First, a new Gini coefficient defined to reduce impact feature redundancy for higher accuracy. Next, number candidate split points calculations continuous features, approximate equal-frequency binning method determine optimal efficiently. Finally, Apache framework, sampling index (FSI) table speed up training process decision trees data communication overhead. Experimental results show that improves constructing forests while ensuring accuracy, superior Spark-MLRF terms performance scalability.