2182. Machine predictive maintenance to prevent defects using Decision tree and Random forest
Invited abstract in session TA-6: Prescriptive Analytics: Sales, Logistics and Industrial Analytics, stream Analytics, Data Science, and Forecasting.
Thursday, 8:45-10:15Room: H9
Authors (first author is the speaker)
| 1. | Pavee Siriruk
|
Abstract
Defect reduction remains a key priority in manufacturing, with the integration of technologies such as the Internet of Things (IoT), machine learning, artificial intelligence, and big data analytics. Predictive maintenance (PdM) has become a critical strategy, using sensor data and advanced algorithms to forecast equipment failures, extend asset lifespans, and improve operational efficiency. The large-scale data collection enhances the accuracy of predictions, enabling timely interventions and reducing the risk of unexpected downtime. In this paper, the factors influencing the occurrence of defects in hard disk drive components are examined using two classification techniques: Decision tree (DT) and Random forest (RF). Three primary datasets are typically split into training and test subsets with ratios of 80:20 and 70:30 to assess model accuracy and ensure reliable predictions, as evaluated using the confusion matrix. Manual and automatic (via GridSearchCV) hyperparameters tuning are applied to optimize the model, thereby enhancing prediction accuracy. The comparison of models across datasets revealed that the random forest model, with manual hyperparameter tuning using data split ratio of 80:20 with Dataset 2, achieved the highest precision and accuracy. An analysis of feature importance was conducted, with the highest score identifying the key factors contributing to the occurrence of the defect under consideration.
Keywords
- Big Data
- Machine Learning
- Forecasting applications
Status: accepted
Back to the list of papers