Composing multi-relations association rules from crowdsourcing remuneration data

Abstract

In crowdsourcing, requesters are companies that require external workers to execute specific tasks, whereas a platform acts as a mediator to match and allocate the tasks to digital workers. To assign it to a worker, the platform must first identify the types of tasks and match them to the appropriate workers based on their level of competency. Each worker has different ICT competencies which affect work quality and remuneration. However, general practise frequently assumes a single level of worker’s capability for all tasks, hence the categorisation of difficulty of tasks is unclear and inconsistent. Apart from causing dissatisfaction among workers, this also implies an absence of incentive standardisation. Therefore, this study explores this matter and which aims to identify and visualise the parameters that affect remuneration determination. To gather the data, focus group discussions and interviews with crowdsourcing players have been conducted. The data comprise a lot of redundancies, therefore an apriori algorithm is used to normalise it by removing redundancies and then extracting significant patterns. Next, an association rule is used to uncover correlations between parameters. To gain a more understandable insight, the data relationship is visualised using an alluvial chart that manages to illustrate the flow. Findings show that task type, outcome variation, and competency requirements demonstrate a degree of interdependence. It is suggested that there is a significant pattern showing that the remuneration scheme is determined by five levels of DW, which are expert, advanced, intermediate, novice, and basic. Advance workers are most likely to participate in the crowdsourcing, and the remuneration scale is suggested to be wider compared to others. The study's findings provide input for remuneration strategy in future work.