|
Classification of offences in narcotic cases using machine learning |
|---|---|
| รหัสดีโอไอ | |
| Creator | Suthat Tanthong |
| Title | Classification of offences in narcotic cases using machine learning |
| Contributor | Waranya Poonnawat |
| Publisher | Mahasarakham University |
| Publication Year | 2569 |
| Journal Title | Journal of Science and Technology Mahasarakham University |
| Journal Vol. | 45 |
| Journal No. | 3 |
| Page no. | 366-377 |
| Keyword | Machine learning, text classification, natural language processing, narcotic cases |
| URL Website | https://li01.tci-thaijo.org/index.php/scimsujournal |
| Website title | Journal of Science and Technology Mahasarakham University |
| ISSN | 1686-9664 (Print), 2586-9795(Online) |
| Abstract | The examination of indictments in narcotic cases requires considerable time and attention to detail, posing challenges for judicial officers. This study aimed to reduce such burdens and enhance the efficiency of the justice process. Therefore, the objective of this research was to develop and evaluate the performance of a machine learning model for classifying offences in narcotic cases. The research consisted of five main steps: (1) collecting 1,651 narcotic case samples from the Suphanburi Provincial Court (January 2022 – May 2024), (2) exploring the dataset, (3) preprocessing the data using PyThaiNLP and TF-IDF, (4) developing a two-stage classification model, where stage one classified narcotic types using a multi-class classification with a One-vs-Rest strategy, and stage two classified the nature of the offence using a multi-label classification with a Classifier Chains strategy, and (5) evaluating the model performance using three algorithms (Support Vector Machine, Logistic Regression, and Random Forest) via 5-fold cross-validation. The results indicated that the Random Forest algorithm achieved the highest performance and stability in both stages, with an F1-score of 99.55% for stage one and 97.70% for stage two, along with a Hamming Loss of 1.58%. These findings demonstrate the potential of the proposed model in assisting the classification of offences in narcotic cases. |
