Advancements in digital grid infrastructure: Exploring automated asset feature identification and disposal

Abstract

In the evolving landscape of digital transformation and intelligent systems, automated asset management is crucial for enhancing operational efficiency and strategic resource planning. This research introduces a novel hybrid model, Logistic Regression-Weighted Random Forest (LR-WRF), combining the benefits of Logistic Regression (LR) and Weighted Random Forest (WRF) to automate asset classification and disposal predictions. The system collects historical data, including asset ID, location, usage history, maintenance logs, and physical condition, which are pre-processed to remove inconsistencies and normalise formats. Principal Component Analysis (PCA) is applied for feature extraction, reducing data dimensionality while retaining essential information. A Weighted Random Forest (WRF) classifier is then used to categorise assets into operational states such as Active, Degraded, or Obsolete. To predict disposal eligibility, a Logistic Regression model assesses whether an asset should be retained, sold, recycled, or discarded. The experimental results show that the LR-WRF model provides high classification accuracy and offers effective disposal recommendations, significantly reducing manual intervention and ensuring compliance with asset lifecycle policies. The proposed model was evaluated across a variety of metrics, including accuracy (98.82%), precision, recall, and F1-score. This research advances the field of intelligent asset management by integrating machine learning techniques into automated decision-making processes, thereby enhancing the efficiency, accuracy, and timeliness of asset management practices in organisations.