Article Preview
Top1. Introduction
In recent decades, crowd-sourcing methodologies have attracted many researchers, because of its effectiveness in real life applications like entity resolution, extracting sentiments from tweets, tagging images, and categorizing product and it enables programmers to integrate with human computations for several tasks. (Borromeo et al., 2017) (de Mattos, Kissimoto, & Laurindo, 2018). Crowd-sourcing aims at soliciting human intelligence for solving machine hard issues and also tackle the tasks like natural language processing, image-tagging and decision making, which are hard for machines, but comparably easy for human workers (Campo, Khan, Papangelis, & Markopoulos, 2018) (Fedorenko, Berthon, & Rabinovich, 2017). Furthermore, crowd sourcing is also utilized to collect the labels for machine learning to perform tasks in computer-human interaction (Berriel, Rossi, de Souza, & Oliveira-Santos, 2017). Several solutions are developed to perform common data-set operations over crowd sourced data like join, count, selection or filtering and sort or rank (Feller, Finnegan, Hayes, & O’Reilly, 2012) (Zhang, Yang, & Liu, 2018). Current crowd-sourcing systems like deco, crowdDB and qurk creates an SQL like query language for declarative interface to the crowd (Bazilinskyy, Kyriakidis, & de Winter, 2015). An SQL like declarative interface is developed in order to encapsulate the complexities for creating a crowd-sourcing system as an interface, which is familiar in large database users. Subsequently, for a given user-query, a declarative system initially compiles the user-query and then develops an execution plan for posting the Human Intelligence Tasks (HITs) to the crowd (Brem, & Bilgram, 2015) (Medury, Grembek, Loukaitou-Sideris, & Shafizadeh, 2017), whereas, the Query Optimization (QO) delivers query interfaces in relational database systems, which is essential for crowd sourcing (Goncalves, Hosio, Liu, & Kostakos, 2016).
The query optimizer finds the most effective way for executing the input query by assuming the possible query plans. Normally, the query optimizer cannot be accessed directly by the operators (Devari, Nikolaev, & He, 2017) (Li et al., 2017). Once the queries are submitted to the database server, parser analyses the queries and then passed to the query optimizer (Acosta, Simperl, Flöck, & Vidal, 2017). The query results are achieved by accessing the relevant data, where the database structures are more complex (Doren, Katherine, Austin, Beth, 2017). In Relational Database Management System (RDBMS), a single data is represented as several copies in multiple data storages. For query optimizer, it is a critical task to derive an optimal plan for answering any user query by retrieving data from RDBMS. The query optimizer identifies the best query plan by means of estimated monetary cost and selection time. In this research work, a multi-objective optimization approach (ant-lion optimizer) is utilized for QO in crowd sourcing. Five major steps followed in ant-lion optimizer are; re-building of traps, entrapment of ants in traps, random walk of ants, catching preys, and building traps. The ant-lion optimizer has high local optima avoidance compared to other evolutionary algorithms. The stochastic operator leads to random changes in the solution, if an evolutionary methodology is surrounded by a local optimum, finally the operator escapes from the local optimum. Ant-lion optimizer is a meta-heurist algorithm, which obtains an optimal shortest path to retrieve the result based on the user query by means of cost, latency and accuracy. This paper is composed as follows. Section 2 survey several recent papers on query-based crowd-sourcing strategies. In section 3, an effective multi-objective optimization methodology: ant-lion optimizer is presented for QO in crowd-sourcing. Section 4 shows comparative experimental result for proposed and existing strategies using UCI automobile and AMT datasets. The conclusion is made in section 5.