Pisit Chanvarasuth

A Virtualization Approach to Auto-Scaling Problem

The paper proposes to simulate feasibility of using virtualization technology to auto-scaling problem in cloud computing. It uses ARENA simulation software to build two different models. There are auto-scaling without server virtualization and auto-scaling with server virtualization. The results of our experiment show that employing virtualization technology increases both life time of servers and CPU utilization.

A Two-Hashing Table Multiple String Pattern Matching Algorithm

Multiple string pattern matching is one of many approaches to simultaneously search occurrences of a large number of patterns in a given text. In this paper, a new solution to this problem is presented by using two-hashing tables to minimize attempting times. This solution, called Two-Hashing Table Multiple String Patterns Matching Algorithm (Two-HT-MSPMA), is suitable for very long length of minimum pattern length. Its time complexity is more efficient than classic algorithms. Empirical results showed that its attempting times were less than of traditional algorithms.

Hybridizing Principles of the ELECTRE III Method with Case-Based Reasoning for a Travel Advisory System: Case Study of Thailand

Despite case-based reasoning (CBR) being one of the collaborative filtering approaches for a travel advisory system, the application of CBR to travel and tourism is still in a very early stage. Although some researches define a similarity function in CBR by using outranking relationship, there is a lack of research on how and why to combine outranking relationship with CBR. In this research, we address how to combine outranking relationship in the ELECTRE III with CBR for a case-based travel advisory system. In order to demonstrate the effectiveness of the hybrid model, the hybrid model is compared to the situation where the non-hybrid model was applied. In travel advisory mining, data were collected from travelers on the Chachoengsao province. Empirical results show that the hybrid model offers significantly better ranking performance than the CBR model derived from Euclidean metric.

A Meiosis Genetic Algorithm

This paper proposes a new genetic algorithm called a meiosis genetic algorithm (MGA) based on meiosis cell division. The meiosis cell division serves producing chromosome diversity which speeds up converging to a mature solution chromosome. The experiment results on a knapsack problem using the new algorithm show that it is more efficient than the simple genetic algorithm.

A Skipping SvS Intersection Algorithm

In this paper, we propose to improve the SvSalgorithm by skipping elements of the smaller set for reducing a search boundary. The boundary of skipped element is determined by boundaries of elements prior to and next to it. We perform experiments on uniformly distributed random datasets to compare our algorithm with the standard SvS.We use regression analysis to get an equation for determining an appropriate skipping number. Our results show that the skipping SvS algorithm using the equation obtained can reduce approximately 41%on the number of comparisons of the standard SvSalgorithm.

The Impact of Business Process Outsourcing on Firm Performance

In recent years outsourcing in general and business process outsourcing in particular has become an important phenomenon. It is widely expected that business process outsourcing offers significant value to firms through lowered costs, increased ability to scale, enhanced flexibility and better access to knowledge and technological resources. However, there are only a few empirical studies investigating the relationships between business process outsourcing and firm performance. This paper attempts to examine the impact of business process outsourcing on the subsequent financial performance of outsourcing client firms. The results indicate that outsourcing intensity has a positive impact on firm productivity and a negative impact on profitability.

A maximal coverage entropy based reduct for classification

The paper proposes a new scoring metric to rank a reduct. Our technique investigates each reduct against the decision table data for ranking by coupling entropy score and coverage score. We compared our work result to a result from a feature selection technique in Weka. The result showed the model learned from our selected features yielded higher accuracy.

A cost and space efficient method for unbalanced assignment problems

In this paper, a new cost and space efficient solution to unbalanced assignment problems is presented. The new solution uses linear space complexity instead of polynomial complexity as used by the Hungarian method. Moreover, its optimal cost is lower than of the Kumar method - an existing linear space complexity solution to unbalanced assignment problems.