Glass recognition as a cutting-edge machine learning approach for identification and classification

Main Article Content

Domaka Nuka Nanwin
Williams Daniel Ofor

Abstract

Crime scene investigations are integral to the justice system, and the efficacy of such investigations can be significantly heightened through precise identification and classification of glass recovered at crime scenes. This research endeavours to address the persistent demand for accurate glass categorization in legal proceedings. The study introduces a comprehensive glass identification and classification model employing a combination of deep neural networks and decision tree algorithms. Utilizing a glass dataset sourced from the UCI machine learning data repository, the research aims to showcase the capabilities and analyze the effectiveness of the developed models. The identification and classification model demonstrated commendable performance, achieving an accuracy of 68% for the training dataset and 60% for the test dataset. The minimal loss of 4.376815 and mean absolute error of 1.412363 underscore the robustness of the proposed model. The investigation employs a variety of tools and test techniques, with a particular emphasis on a confusion matrix table to illustrate misclassifications. Additionally, a tabulated probability distribution of the model provides a nuanced understanding of the percentage distribution in classified data. This research contributes valuable insights into the intersection of forensic science and machine learning, showcasing the potential for advanced technologies to enhance the accuracy of crime scene glass analysis. The findings emphasize the need for continued refinement and development of such models in forensic investigations, paving the way for more sophisticated tools in legal proceedings.

Article Details

How to Cite
Nanwin, D. N., & Ofor, W. D. (2023). Glass recognition as a cutting-edge machine learning approach for identification and classification. Faculty of Natural and Applied Sciences Journal of Scientific Innovations, 5(2), 1–16. Retrieved from https://fnasjournals.com/index.php/FNAS-JSI/article/view/215
Section
Articles