NYCHA Asset Management
PROJECT CONTEXT
The NYCHA Asset Management Project leverages advanced data analytics to improve the decision-making process related to the maintenance and repair of housing assets. By using both historical data and real-time inputs, the project aims to identify spatial trends and key cost factors that influence asset management. This initiative seeks to enhance operational efficiency and cost-effectiveness, ultimately improving the quality of service provided to New York City Housing Authority residents.
METHODOLOGY
Data collection for the project encompasses a broad range of metrics, including the geographic locations of assets, their current states, and detailed repair cost estimations. This comprehensive data acquisition allows for a granular analysis of asset conditions across different neighborhoods.
Traditionally, asset management has relied on manual inspections and assessments based on straightforward criteria such as age or visible wear and tear. While effective to a degree, these methods can overlook subtler, systemic issues that a more data-driven approach can uncover. A machine learning (ML) approach using cluster analysis automates the process of finding similarities and differences in the data to form groups or clusters. ML methods, such as K-means or hierarchical clustering, can process large volumes of data and discover patterns without any prior labeling of the data (unsupervised learning). These techniques are more scalable, can dynamically adapt to new data, and often uncover non-obvious insights that might not be detected through traditional methods.
Traditional Approach
Machine Learning Approach
K-Means Model
This project employs the K-Means Clustering algorithm to organize vast amounts of asset data into meaningful groups. This method optimizes the grouping of assets based on similarity in repair costs and geographical location, allowing us to pinpoint areas of high and low maintenance needs efficiently. By evaluating the variance within each cluster, the algorithm helps in determining the optimal number of groups (K), which is critical for accurate analysis and resource allocation.
Getis-Ord Gi* Method
We utilize the Getis-Ord Gi* method for geostatistical analysis to identify statistically significant hotspots and cold spots of repair costs. This approach not only highlights areas with unexpectedly high or low costs but also ensures that these patterns are statistically robust, rather than random fluctuations. Such insights are vital for prioritizing interventions and optimizing budget allocations.
Together, these advanced methodologies provide a comprehensive and nuanced understanding of asset conditions across NYCHA properties. They facilitate a proactive approach to maintenance, focusing resources where they are needed most.