Improvement of Adaboost Algorithm by using Random Forests as Weak Learner and using this algorithm as statistics machine learning for traffic flow prediction Research proposal for a Ph.D. Thesis

The main goal of this doctoral research is to plan and to build statistic machine learning system for “traffic flow manage control” of urban area for a prediction of traffic flow problem. I also present a new algorithm for improving the accuracy of boosting algorithms for learning binary concepts, which will be used as statistics machine learning. This new algorithm is based on ideas presented in two seminal papers. The first is Schapire (1990), “The strength of weak learn-ability”. The second is Breiman (2001), “Random Forests”. The main idea of the latter is growing an ensemble of trees (it is known that growing many trees instead of just a single tree, results in significant improvements in the classification accuracy). The new algorithm is constructed by combining Random Forests algorithm into Adaboost algorithm as weak learner (Random Forests is a combination of tree predictors, where each tree in the forest depends on the value of some randomized vector θ). This learner is expected to be stronger and more accurate. As I mention above, this new algorithm will be used as tool for prediction of traffic conditions. We assume that better understanding of a traffic system can improve