Assisting Crowdsourced Data Collection by using Data Mining Algorithms.

Crowdsourcing has emerged as an alternative solution to utilize the collective knowledge and expertise of citizen science communities, and it is also considered a much cheaper solution. However, data quality becomes one of the challenges. Lack of motivation and unfamiliarity with labeled data will lead to poor quality of crowdsourced data. This research proposes data mining models to enhance user experience while filling questions in a crowdsourced-based application by building a recommendation system model. The system is implemented on Alboom, an application that provides early warning information regarding Harmful Algae Blooms. The model is built based on the Apriori and DBSCAN algorithms to provide recommendations on responses to the same questions. As a result, users do not need to submit a similar response repeatedly. The Silhouette Coefficient was used to evaluate the clustering model. In addition, we also evaluate the DBSCAN algorithm model by generating and predicting new random locations. The model successfully predicts which cluster it belongs to and outputs ”−1” if the new points are outside the radius of every existing cluster. The recommendation system performance evaluation with live users will be our future research.