MuST 2 -Learn: Multi-view Spatial-Temporal-Type Learning for Heterogeneous Municipal Service Time Estimation Conference: The 33rd ACM International Conference on Advances in Geographic Information Systems; November 3–6, 2025; Minneapolis, MN, USA The 33rd ACM International Conference on Advances in
Nadia Asif, Zhiqing Hong, Shaogang Ren, Xiaonan Zhang, Xiaojun Shang, Yukun Yuan · 2025
Read MuST 2 -Learn: Multi-view Spatial-Temporal-Type Learning for Heterogeneous Municipal Service Time Estimation Conference: The 33rd ACM International Conference on Advances in Geographic Information Systems; November 3–6, 2025; Minneapolis, MN, USA The 33rd ACM International Conference on Advances in by Nadia Asif, Zhiqing Hong, Shaogang Ren, Xiaonan Zhang, Xiaojun Shang, Yukun Yuan (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2508-16503.txt.
Introduction
Nadia Asif and five co-authors present MuST2-Learn, a 2025 framework for guessing how long city service requests will take. Cities run 311 systems where residents report noise, missed trash pickup, and potholes. Callers often hear little about when help will come, which breeds frustration and repeat calls.
The framework reads each request from three angles at once: where it sits, when it arrives, and what type it is. Special parts of the model track links between types and the wide spread of times within one type. On two real data sets it cut average error by at least a third against leading methods, which could make city timelines far more honest.
Catalog record
| CREATOR | Nadia Asif, Zhiqing Hong, Shaogang Ren, Xiaonan Zhang, Xiaojun Shang, Yukun Yuan |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 46 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2508-16503
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
MuST 2 -Learn: Multi-view Spatial-Temporal-Type Learning for Heterogeneous Municipal Service Time Estimation Conference: The 33rd ACM International Conference on Advances in Geographic Information Systems; November 3–6, 2025; Minneapolis, MN, USA The 33rd ACM International Conference on Advances in Geographic Information Systems (SIGSPATIAL ’25), November 3–6, 2025, Minneapolis, MN, USA DOI: 10.1145/3748636.3764173 ISBN: 979-8-4007-2086-4/2025/11 CCS: Information systems Data mining CCS: Information systems Location based services CCS: Applied computing E-government Nadia Asif, Zhiqing Hong, Shaogang Ren, Xiaonan Zhang, Xiaojun Shang, Yukun Yuan Abstract . Non-emergency municipal services, e.g., city 311 systems, have been widely implemented across cities in Canada and the United States to enhance residents’ quality of life. These systems enable residents to report issues, e.g., noise complaints, missed garbage collection, and potholes, via phone calls, mobile applications, or webpages. However, residents are often given limited information about when their service requests will be addressed, which can reduce transparency, lower resident satisfaction, and increase the number of follow-up inquiries. Predicting the service time for municipal service requests is challenging due to several complex factors: (i) dynamic spatial-temporal correlations, (ii) underlying interactions among heterogeneous service request types, and (iii) high variation in service duration even within the same request category. In this work, we propose MuST 2 -Learn: a Mu lti-view S patial- T emporal- T ype Learn ing framework designed to address the aforementioned challenges by jointly modeling spatial, temporal, and service type dimensions. In detail, it incorporates an inter-type encoder to capture relationships among heterogeneous service request types and an intra-type variation encoder to model service time variation within homogeneous types. In addition, a spatiotemporal encoder is integrated to capture spatial and temporal correlations in each request type. The proposed framework is evaluated with extensive experiments using two real-world datasets. The results show that MuST 2 -Learn reduces mean absolute error by at least 32.5%, which outperforms state-of-the-art methods. † † cc-license: by 1. Introduction Non-emergency municipal service systems, such as the 311 platforms, have been widely adopted in cities across Canada and the United States to improve residents’ quality of life ( wang2017structure ; nyc311 ; chi311 ) . These systems allow residents to report local issues, such as noise complaints, missed garbage collection, brush collection, and potholes, through phone calls, mobile applications, or webpages ( nam2014understanding ; hashemi2022automatic ; holmes2007building ) . Once submitted, each service request is forwarded to the appropriate city department and added to its task queue for subsequent processing ( wang2022community ; nam2014understanding ; 10.1145/3463677.3463717 ) . An important function of these service systems is to provide residents with an estimated completion date for their requests, similar to delivery time estimation ( zhao2023hst ) in online shopping and mailing services. Providing estimated service times offers several benefits to both residents and city governments. These include improving government transparency, enhancing resident satisfaction, reducing resident inquiries and staff workload on communication, and supporting more effective planning and allocation of municipal services. Table 1 . City department (division) and its assigned types of requests Department (division) Request type examples Public works (solid waste) Bagged yard waste, brush collection, bulk trash, missed recycle, garbage container repair, missed garbage Public works (street) Loose leaf collection, roadside mowing, potholes, street sweeping, alley maintenance, distressed pavement Public works (forestry) Tree fallen/branch, tree trimming, tree removal, Public works - stormwater Erosion and drainage, flooding, storm drainage problem, storm grates Economic community development Illegal dumping, litter, house violation, abandoned or inoperable vehicle, graffiti removal, zone violation Transportation department Traffic signs, roadway sight obstruction, street lane markings, traffic lights Several studies have investigated city service time and delivery time estimation ( wang2022community ; 10.1145/3286978.3287010 ; raj2021swift ; 9291583 ; zhao2023hst ; zhou2023inductive ; yi2023deepsta ) . ( wang2022community ) applies the Prophet model to estimate the service time of specific 311 service calls by incorporating trend, seasonality, and holiday effects. ( 10.1145/3286978.3287010 ) and ( raj2021swift ) develop structured regression models based on sparse Gaussian Conditional Random Fields to predict future response times. ( zhao2023hst ) proposes a heterogeneous spatial-temporal graph transformer for full-link delivery time estimation in warehouse-distribution systems. In order to estimate package delivery time, ( zhou2023inductive ) designs an inductive graph transformer, which integrates raw feature information and structural graph data. We argue that a key limitation of these works is their treatment of heterogeneous request types as homogeneous, which prevents them from capturing latent inter-type and intra-type correlations that could enhance estimation accuracy. In this work, we aim to integrate the heterogeneity across request types into our learning method to predict the service time of 311 requests. There are several technical challenges of developing such a method. First, service times are not independent across spatial and temporal dimensions. The service time for requests submitted in a given region on a specific day can be strongly influenced by the service times in neighboring regions and previous days. Second, there exist latent interactions among heterogeneous request types. Multiple types of requests are often handled by the same department with limited resources, leading to potential cross-type interference. As a result, the service time of one request type may be affected by the demand level of other types. Third, even within a single request type, service times can vary significantly due to individual request characteristics such as submission time, location, and workload. This intra-type variation reflects the complexity of service time estimation. More analysis results to explain and justify these challenges are provided in Section 2.2 . To address the aforementioned challenges, in this paper, we propose MuST 2 -Learn, a Mu lti-view S patial- T emporal- T ype Learn ing framework, which integrates the latent correlation from the spatial, temporal, and type dimensions. In detail, we first develop an intra-type spatiotemporal encoder using LSTM and CNN to capture latent spatial-temporal patterns for each request type. Second, to model correlations among heterogeneous request types within these patterns, we design an attention-based inter-type encoder that automatically learns the contribution of related types. Third, we introduce an intra-type variation encoder based on Gaussian Process Regression and estimated request workload derived by feeding description of each request into a Large Language Model. This component captures service time variation within each type and the unique characteristics of each request, enhancing prediction robustness. Finally, a multilayer perceptron is adopted to predict service time based on the outputs of the intra-type variation and inter-type encoders. We have engaged with the City of Chattanooga to design a service time predictor and explore its integration into their 311 system, which currently does not provide estimated completion dates for submitted requests. Our goal is to offer this information to residents to enhance transparency and satisfaction. We are in the early stages of deployment, i.e., focusing on feasibility evaluation. In the future, we plan to extend the method and collaborate with additional cities. Our main contributions are summarized as follows. • To the best of our knowledge, this is the first work to perform service time forecasting for municipal service requests considering the heterogeneity across request types and the service time variation of each request type. • We design MuST 2 -Learn: a multi-view learning framework that integrates intra-type spatiotemporal and variation encoders with an inter-type heterogeneity encoder to capture spatial, temporal, and type-specific