Application of Machine Learning for Optimization

Introduction

The pursuit to make intelligent machines that can coordinate and conceivably rival people in thinking and settling on savvy choices returns to at minimum the beginning of the improvement of the computerized registering in the last part of the 1950s (Bennett and Mangasarian, 1993). The objective is to empower the machines to fill complex problem roles by gaining from the previous encounters and afterward taking care of the perplexing issues under conditions that are fluctuating from the past perceptions. Machine learning is quickly developing, with numerous hypothetical logical advances and applications in an assortment of fields. Analysts have concentrated on improvement as a significant piece of Artificial Intelligence. The common purpose is to delegate machines to accomplish the cognitive functions by learning from the past occurrences and then solving the complex problems under the state of affairs and different forms of the past consideration. The optimization calculation is built upon an emphasis conspire that proceeds to select the modern input values so that the most extreme or least of the objective work is accomplished (Rousu et al, 2006). Driven by the exponential growth of computing techniques and the data collection and a wide range of practical applications, machine learning is now a strategically important area in the field of optimization algorithm method (Nemhauser and Wolsey, 1999). The recognizable proof of the basic plan, design factors, which are dominatingly changed during the advancement cycle of mathematical optimization process and programming model, is the initial phase in forming an optimization problem (Rousu et al, 2006). In Data fitting, optimization algorithm is built upon a cycle plot that proceeds to pick out latest input values so that the most extreme or least value of the proficient objective data functions. Many new algorithmic, theoretical, mathematical and computational benefaction of optimization have been proposed to decipher several data function problems in utilization of Machine Learning mechanism (Flaxman et al, 2005). Artificial Intelligence depends on the improvement of a model that gives the fruitful result when given a particular information. The heuristic method has been initiated further supple and systematic structured than the deterministic viewpoint. Though the obtained data solution quality ineffective or not guaranteed (Flaxman et al, 2005).

Machine learning approaches have been increasingly prominent in the optimization methodologies in recent years. To help machine learning algorithms discover faster answers, several optimization strategies have been developed. The gradient descent approach, for example, is a common optimization technique for quickly finding the optimal weight sets. The goal of this book chapter is to include both the original research papers and review papers on many disciplines of machine learning applications, with an emphasis on the optimization approaches and optimization algorithms for the time-series data analytics. Optimization methodologies take part in a dominant role in machine learning projects as well as adapting the learning algorithms for training of datasets. The step of preparing the data before fitting the model and the step of adjusting the selected model are also called optimization problems. In fact, the entire predictive modelling approaches can be seen as one big optimization problem as a whole. In this chapter, we show the interaction between machine learning and optimization method and also we examine how many types of algorithm have in optimization technique and how they are related to each other. Both the methodologies play a crucial role in customer support analytics, especially in E-commerce websites.

There are some brief applications where machine learning algorithms are used and optimization techniques plays an important role to optimize the solutions.