Data Analysis and Applications 3, 3rd Edition

7Support Vector Machines: A Review and Applications in Statistical Process Monitoring

Modern industrial problems have become increasingly complex, and classical process monitoring techniques do not suffice to solve them. Hence, statistical learning methodologies have nowadays become popular in this area. In this chapter, we examine a specific statistical learning technique using support vector machines. It is the most powerful algorithm used in different fields of statistics and computer science. We present a review of the literature concerning support vector machines in the process monitoring field, test one of the mentioned works on a real data set and, finally, present an alternative approach which is able to yield better results.

7.1. Introduction

The term statistical process control refers to a wide variety of statistical tools used to improve the performance of a process and ensure the quality of the products produced. The main use of statistical process control is in industrial environments, but this is not always the case, since many techniques are used in financial or other kinds of problems. In practice, the way statistics is used in quality control is through the construction of control charts (see, for example, [REY 90, HAW 05]).

In order for a statistician to check whether a process is in or out of control, a statistic is calculated using samples taken from the process, and, if it takes a value outside of some specified control limits, then the process is out ...