Quantity forecast of administrative items based on parallel random forest

Abstract

The ultimate goal of this paper is to train a model based on the given administrative data to predict the amount of each administrative item of month in different years and different regions as accurate as possible. In this paper, we propose a novel approach for quantity forecast of administrative data which is named after parallel random forest (parallel RF). Firstly, we collect administrative data from different online systems using java program and store it in MongoDB. Then we extract key information from these data and assign different numbers to different administrative areas and item names. Next, as the core of whole method, we train the prediction model by implementing the random forest method on Hadoop Map-Reduce. Finally, we compare the execution efficiency and prediction accuracy with other standard algorithms such as SVM and gradient boosting. The experiment shows that the accuracy and efficiency of our method is much better than other algorithms and our method is more reliable and useful.