Automating Machine Learning Data Preprocessing Pipelines

Nicolas Carrasco

Bakalářská práce

Abstract:

Data preprocessing is a particularly important phase of machine learning workflows (ML). However, it always ends up being time-consuming, often requiring a significant amount of manual effort and domain ability. For example, improper handling of the data could lead to misleading results and hinder the reproducibility of the model. The aim of this thesis is to design an automated machine learning data processing framework, to then implement and evaluate it, focusing on efficiency, accuracy, and reproducibility. The approach taken is a modular Python-based preprocessing pipeline, built using the library Scikit-learn, more specifically the Pipelines module. The framework automates the common data preparation tasks, like outlier handling, imputation, feature scaling, categorical encoding, and optional feature generation (featuretools). This design aims to be used across different datasets, also supporting further experimentation. The evaluation of the effectiveness of the automated preprocessing framework is through a comparative study, using real-world datasets, more specifically, ore excavation. The comparison between the automated approach and the manually implemented preprocessing framework will be measured in terms of execution time, computational resource usage, and model performance using the Root Mean Squared Error (RMSE). The findings show that the usage of automated preprocessing pipelines does stand for a more practical and effective solution, however, always including supervision of what is being done.

Keywords

Preprocessing pipeline, Data preparation, Experimentation, Comparative study, Execution time, Manual effort, Computational resources, RMSE

Jazyk práce: angličtina

Datum vytvoření / odevzdání či podání práce: 17. 3. 2026

Obhajoba závěrečné práce

• Obhajoba proběhla 19. 5. 2026
• Vedoucí: Ing. Jakub Konopásek, Ph.D.
• Oponent: Vojtěch Pivoňka, Ing.

Česká zemědělská univerzita v Praze

Provozně ekonomická fakulta

Bakalářský studijní program:
Informatics