Crowdsourcing platforms are vulnerable to spam and low-quality submissions, undermining data integrity. Effective spam detection is crucial. This paper presents a novel approach using ensemble attribute selection techniques to improve spam classification in crowdsourcing environments. Traditional methods rely on predefined features that may not capture evolving spam patterns. Our method employs an ensemble of attribute selectors to identify the most informative features from crowdsourced data. By combining multiple strategies, it can effectively identify spam-indicative attributes. Evaluated on real-world crowdsourcing datasets, our ensemble approach outperforms traditional techniques in accuracy, precision, and recall. We analyse selected attributes to provide insights into spam characteristics in crowdsourcing. Contributions include: (1) a robust spam classification method for crowdsourcing and (2) valuable understanding of crowdsourced spam nature. Our findings enhance crowdsourced data quality and reliability, benefiting applications leveraging this resource. The ensemble attribute modelling approach effectively tackles the spam challenge in crowdsourcing platforms. With comparison to other state-of-the-art method, the proposed ensemble approach for attribute selection performs better.