An automated prescriptive domain data preprocessing algorithm to support multilabel‐multicriteria classification for Indian coastal dataset, crop dataset, and breast cancer dataset

SummaryIn many real‐world applications, the problem of machine learning is based on multicriteria. The multicriteria learning process becomes problematic with data classification involving multiple class labels. The multilabel‐multicriteria classification (MLMCC) is an emerging research area among the data science community, and it has applications in many domains like sciences, computer vision, government, business, and engineering domains. Most of the existing classification techniques learn single criteria and fail to focus with the multicriteria‐based internal information. In order to overcome the above research problem, we propose an expert‐based pattern driven multilabel learning algorithms (E‐PDMLAs) to handle MLMCC. The whole framework is supported with expert knowledge observed as multicriteria, and it is summarized into three broad categories of linguistic terms, namely, MoreThan(>=), LessThan(<=), and Between(‐). Also, the top level of data analytics is to unfold the prescriptive analysis on the data. In this work, prescribing data is supported with linguistic terms, namely, Critical, Feasible, and Strong. In support of E‐PDMLA, the data have been preprocessed and prescribed into linguistic terms for the better insight of information. In this paper, we propose a novel automated prescriptive and fuzzy‐based preprocessing algorithm (prescriptive fuzzy‐based domain data preprocessing technique [PF‐DDPT]) for all the above three categories. The experimental study has been carried out with Indian coastal dataset, crop dataset, and breast cancer dataset, and the results have been compared with normal and modeling technique. The analysis manifests the strength of our proposed algorithm (PF‐DDPT) over accuracy on preprocessing and prescription based on expert perception.