Relating mathematics to machine learning through algorithm development for development for big data analysis

Abstract

Data has increased at an exponential rate and has outpaced our capability to analyze it. However, new ways of data analysis, which thrive in big data such as Machine Learning (ML) have emerged. This study explores Machine Learning by creating a Machine Learning algorithm based on Support Vectors. This was done by converting mathematical formulations into a computer algorithm that was then used for data classification. The algorithm was evaluated and compared to other algorithms. The results of the evaluation show that the algorithm was accurate at binary classification. Comparisons to other algorithms using both the iris and breast cancer datasets show that algorithms based on Support Vectors are generally more accurate at data classification. This means that the approach that was used in this study can be used in businesses to determine whether a person will return loan or not or whether a particular student can finish a degree program or not based on past data. The study also indicated that Support Vector Machines algorithm training require more computing power as data gets bigger. Hence, it suggested use of high performance computing for big data analysis.