PROJECT TITLE :

Cluster-Based Boosting

ABSTRACT:

Boosting is an iterative method that improves the predictive accuracy for supervised (machine) learning algorithms. Boosting operates by learning multiple functions with subsequent functions that specialize in incorrect instances where the previous functions predicted the wrong label. Despite considerable success, boosting still has difficulty on knowledge sets with bound varieties of problematic training data (e.g., label noise) and when advanced functions overfit the training knowledge. We propose a completely unique cluster-based boosting (CBB) approach to address limitations in boosting for supervised learning systems. Our CBB approach partitions the training knowledge into clusters containing highly similar member data and integrates these clusters directly into the boosting process. CBB boosts selectively (using a high learning rate, low learning rate, or not boosting) on each cluster based on each the extra structure provided by the cluster and previous function accuracy on the member knowledge. Selective boosting allows CBB to boost predictive accuracy on problematic coaching information. Additionally, boosting separately on clusters reduces function complexity to mitigate overfitting. We have a tendency to give comprehensive experimental results on 20 UCI benchmark knowledge sets with 3 completely different kinds of supervised learning systems. These results demonstrate the effectiveness of our CBB approach compared to a common boosting algorithm, an algorithm that uses clusters to boost boosting, and two algorithms that use selective boosting while not clustering.