Data mining of school inspection reports to identify the assets with top renewal priority

Abstract

Building components require frequent inspections to help allocate renewal funds to the most needy assets. However, the low granularity of inspections puts many assets at the same priority level, thus leaving asset managers with difficult decisions as to which assets are more deserving of the limited funds available. This study aims to improve fund-allocation decisions by applying data mining techniques to provide a more granular classification of asset criticality. The study relies on textual information from inspection reports of 400 schools across Toronto and focuses on roofings due to their higher need for repairs. First, rule-based text mining is used to identify the damage type and extent from the damage descriptions in the inspection reports. Then, unsupervised clustering is used to classify the suggested renewal events into four criticality levels based on the extracted data and other relevant parameters such as age. The proposed framework could identify that only 8.8% of deteriorated roofs are highly critical and can be funded with the available limited budget. The proposed approach is adaptable to other asset types and components and can substantially improve the fund allocation practice for large owner organizations such as municipalities and school boards.