Civic Link: An AI-Driven Hybrid Recommendation and Workload-Aware Resource Allocation System for NGO - Community Collaboration

Abstract

National and international non-governmental organizations (NGO’s) are instrumental in providing essential social services, including food distribution, healthcare provision, support to education and community empowerment. The increased popularity of digital support platforms that are used to match donations and volunteers with community demand, has made it much harder for requests being placed by communities in need to be matched up with those who can help them. Traditional NGO coordination mechanisms are mainly based on static listing and manual matching, which lead to donation mismatching, uneven workload among the NGO’s and time delay in response. Also, existing platforms do not provide evaluation for the quality of donation descriptions neither predict the workload capacity an NGO can accept before a donation is performed.

To overcome the limitations, we proposed CivicLink in this paper, which is an AI-based Hybrid recommendation and workload-aware resource allocation system to bridge NGO-community. The proposed mechanism is using a text-based NLP classifier in order to automatically determine the relevance and usability of donation or service requested. A hybrid NGO recommendation engine that utilizes content-based filtering, collaborative-filtering (CF), as well as geospatial clustering is then used to recommend the most suitable NGO’s based on specialization, past collaborative history and geographical proximity. Also, prediction of NGO workload is performed based on regression models and time-series forecasting algorithms and used for proactive capacity-aware decision making. Metaheuristic optimization algorithms are used to resolve the allocation problem when there are several prospective NGO’s.

Experimental results under synthetic conditions show that CivicLink outperforms several performance criteria including matching accuracy, work load balancing and response time in contrast to classical NGO’s coordinating patterns. The experimental results demonstrate the feasibility of unifying verification, recommendation, prediction and optimization over a single decision-support framework, which makes CivicLink applicable for real-world civic management applications.