•  
  •  
 

Corresponding Author ORCID Identifer

Choo-Yee Ting https://orcid.org/0000-0001-5667-2816
Thanh-Trung Nguyen https://orcid.org/0009-0000-9948-1753

Section

Regular Issue

Abstract

Data preprocessing is a foundational stage in any machine learning (ML) workflow, especially when dealing with complex, multilingual, and real-world healthcare datasets. The quality, consistency, and structure of input data strongly influence the accuracy, fairness, and interpretability of predictive models. This paper presents a comprehensive and structured preprocessing framework designed for a large multilingual healthcare dataset containing 19,536 patient records with 67 features encompassing demographic, clinical, and socioeconomic information. The framework systematically addresses data heterogeneity, linguistic diversity, and structural inconsistencies through multiple stages of cleaning and transformation. The preprocessing process includes translation of Malay text into English to ensure linguistic uniformity, standardized handling of missing values using transparent placeholders, and column splitting of composite attributes to improve granularity and feature representation. Outlier detection was performed using statistical approaches such as Interquartile Range (IQR) and Z-Score methods to identify and remove extreme or erroneous values, thereby improving data integrity and model stability. Furthermore, categorical variables were converted into numerical form using one-hot encoding, ensuring compatibility with diverse ML algorithms while preventing artificial ordering and bias. An experimental evaluation using a Random Forest classifier demonstrated that these preprocessing techniques substantially enhanced predictive performance. Model accuracy improved from 72.4% to 86.7%, while the area under the curve (AUC) increased from 0.78 to 0.92, indicating more robust classification and generalization. The resulting dataset is cleaner, more consistent, and better suited for downstream ML applications. Overall, this study provides a reproducible and practical framework for managing multilingual healthcare data, underscoring that systematic preprocessing is not merely a preliminary task but a critical foundation for building reliable, transparent, and interpretable machine learning solutions in healthcare and other data-intensive domains.

Download

Share

COinS