In order to make the data more manageable and improve the accuracy of machine learning models, data need to be preprocessed. Data preprocessing involves a series of steps such as data cleaning, data transformation, feature selection, and feature extraction. Data cleaning involves removing or correcting data that is incorrect, incomplete, irrelevant, or duplicated. Data transformation includes normalization, standardization, aggregation, and discretization. Feature selection involves selecting the most relevant features from the dataset while feature extraction involves transforming existing features into new features. Our proposed evaluation function combines a multi-label accuracy measure with a feature selection measure. In this paper a new assessment method based on swarm intelligence for feature selection. The suggested technique utilizes a multi-objective fitness function by utilizing mutual information between features and labels. We conduct experiments on five real-world datasets. The results show that proposed evaluation function in some cases outperforms the existing methods.