Evaluation of Statistical Learning Methods for FFF 3D Printing Problems
- 31 July 2026
- journal article
- Published by Trans Tech Publications, Ltd. in Key Engineering Materials
- Vol. 1060, 71-78
Abstract
This study investigates the applicability of several statistical learning regression and classification methods for analyzing fused filament fabrication (FFF) printing parameters and their relationship to resulting mechanical properties reported in different papers. The research questions are the following. RQ1: Identify which algorithms perform best on relatively small dataset with moderate dimensionality. RQ2: Investigate that secondary data extracted from publications are sufficiently uniform for reliable modeling, considering the potential influence of unreported printing parameters, environment variables. Data were collected using a python-based extraction process to retrieve data from the literature, followed by manual refinement, cleaning. Feature selection and engineering were applied to standardize input features and address missing datapoint values, resulting in a dataset of approximately 500 samples with eight input features and three output mechanical properties. Various regression methods were tested like linear regression with regularization, local regression, random forest regression, and gradient boosting regression, while classification methods include logistic regression, random forest classifier, gradient boosted trees, and support vector machines. Models were examined using standard regression and classification metrics, including residual analysis, confusion matrices, and cross-validation. Results show that missing features significantly influence model performance. The study shows the challenges and potentials of applying machine learning to secondary data in additive manufacturing.Keywords
This publication has 7 references indexed in Scilit:
- Dataset meta-level and statistical features affect machine learning performanceScientific Reports, 2024
- Academic Insights and Perspectives in 3D Printing: A Bibliometric ReviewApplied Sciences, 2021
- Model Loss and Distribution Analysis of Regression Problems in Machine Learning2019
- Multicollinearity and Regression AnalysisJournal of Physics: Conference Series, 2017
- Statistical Models2009
- Review: A gentle introduction to imputation of missing valuesJournal of Clinical Epidemiology, 2006
- The predictive accuracy of artificial neural networks and multiple regression in the case of skewed data: exploration of some issuesExpert Systems with Applications, 2000