Sylvian-Silva (SFORCE): An Ensembled Boost Approach Towards Machine Learning
Anand Ravishankar, Santhi Natarajan, Bharathi Malakreddy A · 2020
Gathering a minimalistic set of convincing evidence for a logical conclusion formed the premises of Occam's Razor, put forth by Libert Froidmont. Occam's Razor, although sound and appealing for theoretical proofs, is speculative and vague when it comes to practical scientific experiments with multivariate big data sets. The advent of the "Principle of multiple explanations" forms the basis of all ensemble architectures for Machine Learning. Random forests are an ensemble of decision trees that provide excellent results among its competitors. The main disadvantages of a random forest are the computational cost and the curse of overfitting. This paper presents Sylvian Silva (SFORCE), an Ensembled Boost approach towards Machine Learning that combines the strength of Random Forest ensemble and the boosting approaches. Stagewise Additive Modeling and real-valued confidence related predictions for Multiclass classification and regression form the basis of these boosting algorithms. Boosting ensures that the algorithm gathers the inherent characteristics and has the capability to predict future targets. The SFORCE algorithm's performance is compelling across six different data sets. SFORCE reports high accuracy levels consistently without overfitting when compared with 17 different machine learning algorithms for classification and 10 different algorithms for regression.