Pith. sign in

REVIEW 3 major objections 5 minor 1 cited by

Neurosymbolic AI for Travel Demand Prediction: Integrating Decision Tree Rules into Neural Networks

T0 review · 3 major / 5 minor · reviewed 2026-08-09 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read This paper claims that feeding binary decision-tree-rule features into a neural network improves travel-demand prediction accuracy and commuter-pattern alignment beyond using raw geospatial, economic, and mobility features alone.

desk verdict A clean application of rule-feature augmentation, but the missing train/test split and target-derived rule features make the reported gains uninterpretable. read the letter →

arxiv 2502.01680 v1 pith:S55C5VPZ submitted 2025-02-02 cs.LG cs.AI

Kamal Acharya , Mehul Lad , Liang Sun , Houbing Song This is my paper · ORCID
classification cs.LGcs.AI
keywords neurosymbolicAItraveldemandpredictiondecisiontreerulesneuralnetworksfeatureaugmentationvariance-basedruleselectioncommuterflowinterpretablemachinelearning
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper argues that symbolic rules extracted from decision trees, when appended to a neural network's input as binary features, improve county-to-county travel-demand prediction in Tennessee across mean absolute error, R-squared, and the Common Part of Commuters metric. The authors build a dataset from mobility, demographic, economic, and geographic sources, train decision trees of depths 3 through 15, convert every root-to-leaf path into an if-then rule, and filter rules by variance thresholds of 0.01, 0.001, and 0.0001. They report that the combined dataset (original features plus rule indicators) consistently outperforms both the raw features alone and the rules alone, with the finest variance threshold (0.0001) giving the strongest gains at larger tree depths. The paper frames this as a neurosymbolic recipe that preserves interpretable structure while retaining neural accuracy.

What carries the argument

The central mechanism is rule-as-feature augmentation. A decision tree of depth $d$ partitions the feature space into leaves; each root-to-leaf path is converted into an if-then rule, and each rule becomes a binary indicator feature (1 if the rule's conditions are satisfied for that origin-destination pair, 0 otherwise). These binary rule features are appended to the original 11-column dataset, and a neural network is trained on the combined feature set. Variance-based rule selection (excluding rules with variance below 0.01, 0.001, or 0.0001) controls how many of these binary features survive, trading off rule quantity against granularity. The paper's claim is that this specific integration—not the neural network alone and not the rules alone—produces the accuracy gains.

What would settle it

Train the decision trees and the neural network on a proper cross-validation split (e.g., extract rules only from training folds, then evaluate on held-out folds) and compare the combined dataset's MAE, R-squared, and CPC against the baseline on the held-out data; if the combined dataset no longer beats the baseline, leakage is confirmed.

Watch

Extended reading notes

Core claim

The central claim is that a hybrid model—a neural network whose input includes binary indicators for decision-tree rules—predicts county-to-county travel flows more accurately than the same network trained on raw geospatial, economic, and mobility features alone. Across decision-tree depths 3 through 15, the combined dataset achieves lower MAE, higher R-squared, and higher CPC than either the final dataset alone or the rule-only datasets, with the advantage growing at deeper trees. Among variance-based rule filters, the 0.0001 threshold retains more low-variance rules and gives the best results on all three metrics, which the paper attributes to capturing finer-grained patterns. The paper presents this as a general neurosymbolic approach: extract interpretable if-then rules with a decision tree, encode each rule as a binary feature, and let the neural network combine them with the raw features.

Load-bearing premise

The decision trees that generate the rules are trained on the full dataset including the flow values being predicted, and the neural network is then trained and evaluated on that same dataset, with no described train/test split, so the reported rule gains may reflect information leakage rather than genuine predictive skill.

Editorial extensions

If this is right

  • If the reported gains hold, practitioners can improve travel-demand forecasts simply by appending decision-tree-rule indicators to existing tabular feature sets, without changing the neural architecture.
  • Finer variance thresholds (0.0001) should be preferred when enough rules are available, since they preserve the low-frequency patterns that drive commuter overlap.
  • Deeper decision trees (depth 10-14) generate thousands of rules but deliver diminishing returns; variance filtering is what keeps the rule set tractable and useful.
  • The extracted rules provide directly inspectable, human-readable statements about what drives flows (e.g., short distance and high destination POI counts), supporting the interpretability claim.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Because the decision trees are trained on the full dataset including the flow target, and the neural network is trained on that same dataset, the reported gains may partly reflect information leakage rather than genuine predictive generalization; a proper train/test split with rule extraction performed inside cross-validation folds would test this.
  • The rule-as-feature idea is a special case of hypothesis-space augmentation; any interpretable model (e.g., logistic regression, rule lists) could replace the decision tree, and the comparison would reveal whether the gain comes from the tree's specific partitions or from the extra features generally.
  • The CPC metric rewards alignment of total origin-destination flow patterns, not per-pair accuracy; the gains in CPC may be driven by improved ranking of large flows, which is exactly what the rules at 0.0001 variance capture (low-frequency but high-flow pairs).
  • A natural testable extension is to apply the same rule-augmented network to other regions or to temporal holdouts (e.g., train on March, test on April) to see whether the advantage persists outside Tennessee in the same period.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. The paper proposes a neurosymbolic travel-demand prediction framework for county-to-county flows in Tennessee. It trains decision trees of depths 3–15 on the full feature set and the target flow variable, extracts if-then rules from tree paths, encodes those rules as binary features, filters the rule features by variance thresholds, and feeds them together with the original features into a neural network. The authors report that the rule-augmented 'Combined Dataset' outperforms the baseline 'Final Dataset' and the rule-only dataset across MAE, R², and CPC, with the finest variance threshold (0.0001) giving the best results. The paper includes a methodology description, three sets of experimental figures, a rule-count table, a discussion of limitations, and a link to GitHub code.

Significance. If the claimed gains are real, the idea of using decision-tree-rule features as inputs to a neural network is a simple and potentially practical way to inject interpretable structure into a demand-prediction model. The paper contributes a concrete application of neurosymbolic ideas to a transportation domain where such hybrid approaches are not yet standard, and it provides code and data access, which is a strength. However, the current evaluation does not establish the central claim: all reported results appear to be in-sample, with rule construction and neural-network training and evaluation happening on the same rows, and no error bars or significance tests are reported. As a result, the contribution is currently an interesting proposal rather than a validated method; a proper held-out evaluation is required before the performance comparisons can be assessed.

major comments (3)
  1. [Sections III-B, III-D, IV-B] No train/test split, temporal holdout, or cross-validation is described. The decision trees in Section III-B are trained on the full dataset with the target flow variable; Section III-C applies the extracted rules to the same rows and filters them using variance computed on those rows; Section III-D trains the neural network on these rule-augmented features; Section IV-B reports MAE, R², and CPC on the same data. Because the rule conditions and variance-based selections are derived from the target, the evaluation is in-sample, and the reported improvements may be entirely a result of label leakage. The central claim that the Combined Dataset outperforms the Final Dataset is therefore unsupported. The experiments must be redone with a protocol in which rule extraction and variance-threshold selection are performed inside each training fold and metrics are reported only on held-out data. The limitation section (V-D) discusses scalability and threshold bias but does not acknowledge this leakage risk, which is the most important missing limitation.
  2. [Section IV-B] The comparisons across tree depths and variance thresholds are reported without error bars, repeated runs, or significance tests. Section III-D does not specify the neural-network architecture, number of epochs, optimizer, learning rate, regularization, or any other training detail, so the results are not reproducible and the observed differences could be within-run variability. In addition, the 0.0001 threshold is identified as best after inspecting the same results, so the threshold selection is not protected against overfitting to the evaluation set. The authors should pre-specify the threshold or select it on a validation set and report variability over multiple random seeds.
  3. [Section IV-A3] The CPC computation is not operationalized. Equation (3) requires generated flows y_g, but the paper never states how continuous neural-network predictions are converted into the generated flows used in the CPC index, for example by thresholding, rounding, or proportional assignment. Without this definition, the CPC values in Figures 3c, 4c, and 5c cannot be reproduced or interpreted, and the claim of superior alignment with observed commuter patterns is not verifiable.
minor comments (5)
  1. [Table III] The table lists 12 datasets for each of the rule-based configurations with DT depths 3–15, but the inclusive range 3 through 15 contains 13 depths. The counts should be 13 (or the depth range should be stated correctly), and the statement in Section III-D that 60 distinct rule-derived datasets are generated should be checked accordingly.
  2. [Section III-C] The filtering direction is ambiguous: the text says rules with variance falling below thresholds are excluded, while the Table IV header 'Rules Selected by Variance' suggests the opposite. Please clarify whether low-variance rules are kept or dropped and define how variance is computed on the binary rule indicators.
  3. [Sections IV-B and V] The text describes the figures only qualitatively, with no numeric values, error bars, or axis ranges reported in any table. The magnitude of the claimed improvement over the baseline cannot be assessed from the current presentation.
  4. [Section I] The statement that 'no prior studies have investigated the application of Neurosymbolic AI in the domain of travel demand modeling' is a strong claim that would require a systematic literature search to support; it should be softened or backed by explicit evidence.
  5. [Section III-A] The description of the initial dataset says there are 22 columns for origin and destination features, 2 for time and distance, 1 for weekday, and 1 for flow, but the text later says the final dataset contains 10 features plus the flow column. Listing the retained features explicitly would improve reproducibility.

Circularity Check

1 steps flagged · score 7.0 of 10

Central claim reduces to target-derived DT rule features evaluated on the same dataset; no train/test split described.

  1. fitted input called prediction [Sections III-B, III-C, III-D and IV-B]
    "We employed DT as a method to derive interpretable rules that elucidate the relationships between features and target variables. ... After training the DTs, the paths from root nodes to leaf nodes were traversed to extract if-then rules. ... The extracted rules from the DT were then applied to the dataset to generate a new set of features. ... The Figure 3 illustrates the performance of the NN model evaluated on three datasets: the Final Dataset, the Rules Dataset, and the Combined Dataset (Rules + Final Dataset)."

    The decision trees are trained on the full dataset with the flow variable as the target, so the extracted if-then rules are fit to the label column. Those same rules are then applied to the same rows to create binary rule features, and the NN is trained and evaluated on that same data; no train/test split, temporal holdout, or cross-validation is described anywhere in Sections III or IV. Since the rule features are target-derived indicators, the NN can use leaf membership on a training row to recover label information encoded in the splits, making the reported improvements of the Combined Dataset over the Final Dataset in-sample comparisons rather than independent predictions.

full rationale

The paper's central empirical claim is that augmenting the final dataset with decision-tree-rule features consistently improves travel-demand prediction. The derivation chain, however, is not self-contained against a clean evaluation. The DTs are explicitly constructed to 'elucidate the relationships between features and target variables,' and the extracted rules are then 'applied to the dataset'—the same dataset used to train those DTs—to generate binary features. The NN is trained and 'evaluated' on that same data, with no split described. This is the fitted-input-called-prediction pattern: the rule features encode information from the target variable, and the reported MAE/R2/CPC numbers are therefore at least partly in-sample fits, not predictions on held-out rows. The variance-based rule selection does not cure the problem; lower thresholds keep more target-informed rules, which explains the observed monotonic benefit and makes the comparison circular in effect. The paper's self-citations to prior work ([10], [11], [14]) are related-work context and are not load-bearing, and there is no imported uniqueness theorem or ansatz-by-citation issue. The limitations section acknowledges scalability and threshold bias but never addresses evaluation leakage. Because the central claim reduces to target-derived features evaluated on the same rows, the circularity score is 7 rather than lower. A proper temporal or random holdout evaluation could break the circularity and support the claim, but the manuscript as written does not provide one.

Assumptions & free parameters 3 free parameters · 3 assumptions · 0 invented entities

The paper introduces no invented physical entities. The free parameters are the rule-selection thresholds, tree depths, and unreported NN hyperparameters. The main structural assumption is that in-sample rule extraction does not leak target information.

free parameters (3)
  • variance threshold for rule selection = 0.01, 0.001, 0.0001
    Three thresholds are tested and the best-performing threshold (0.0001) is highlighted in the conclusions; these are hand-chosen, not derived.
  • decision tree depth = 3 to 15 (swept)
    Depths are swept to find the best results; the choice of best depth is post hoc.
  • neural network hyperparameters = not reported
    Architecture, layer sizes, learning rate, epochs, regularization, and optimizer are not stated, but they affect reported performance.
assumptions (3)
  • ad hoc to paper Rules extracted from decision trees trained on the full dataset can be used as features without leaking target information
    The workflow trains DTs on the full data and then trains/evaluates the NN on the same data (Section III-B and IV-B). No holdout split is described, so this assumption is necessary for the reported metrics to be unbiased.
  • domain assumption The SafeGraph-derived mobility data is a valid ground truth for county-to-county commuter flows
    The paper uses the multiscale human mobility flow dataset as the target without validating it against census or survey-based commuting flows (Section III-A).
  • domain assumption County-level features and one month of daily flows are sufficient to model travel demand
    The feature set is static and limited to Tennessee; temporal dynamics are acknowledged as a limitation in Section V-D.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Neurosymbolic AI for Travel Demand Prediction: Integrating Decision Tree Rules into Neural Networks." pith.science (2026). https://pith.science/paper/S55C5VPZ

@misc{pith2026250201680,
  author       = {Pith},
  title        = {Pith review of: Neurosymbolic AI for Travel Demand Prediction: Integrating Decision Tree Rules into Neural Networks},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/S55C5VPZ}},
  note         = {Machine review of arXiv:2502.01680}
}
abstract

Travel demand prediction is crucial for optimizing transportation planning, resource allocation, and infrastructure development, ensuring efficient mobility and economic sustainability. This study introduces a Neurosymbolic Artificial Intelligence (Neurosymbolic AI) framework that integrates decision tree (DT)-based symbolic rules with neural networks (NNs) to predict travel demand, leveraging the interpretability of symbolic reasoning and the predictive power of neural learning. The framework utilizes data from diverse sources, including geospatial, economic, and mobility datasets, to build a comprehensive feature set. DTs are employed to extract interpretable if-then rules that capture key patterns, which are then incorporated as additional features into a NN to enhance its predictive capabilities. Experimental results show that the combined dataset, enriched with symbolic rules, consistently outperforms standalone datasets across multiple evaluation metrics, including Mean Absolute Error (MAE), \(R^2\), and Common Part of Commuters (CPC). Rules selected at finer variance thresholds (e.g., 0.0001) demonstrate superior effectiveness in capturing nuanced relationships, reducing prediction errors, and aligning with observed commuter patterns. By merging symbolic and neural learning paradigms, this Neurosymbolic approach achieves both interpretability and accuracy.

Figures

Figures reproduced from arXiv: 2502.01680 by the authors.

Figure 1
Figure 1. Neurosymbolic Approach Used in the Research [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Methodology of the Research TABLE I: Datasets Used in the Research Name Details Mobility Dataset Multiscale Dynamic Human Mobility Flow Dataset [23] Population Dataset Tennessee State Data Center [24]. Distance and Time Between Counties Open Source Routing Machine (OSRM) [25] Geographical features Open Source Routing Machine (OSRM) [25] Economic features Tennessee Economic Indicators [26] Tennessee County Profiles T… view at source ↗
Figure 3
Figure 3. NN performance on Final, Rules and Combined dataset [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: NN performance on various sets of rules selected on the basis of variances [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: NN performance per rules on different datasets [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Integrating Neurosymbolic AI in Advanced Air Mobility: A Comprehensive Survey

    cs.RO 2025-08 conditional novelty 4.0 of 10

    A survey mapping how neurosymbolic AI could address safety, regulatory, and operational challenges in advanced air mobility.

Reference graph

Works this paper leans on

27 extracted references · 24 canonical work pages · cited by 1 Pith paper

  1. [1]

    Passenger demand fore- casting in scheduled transportation,

    N. Banerjee, A. Morton, and K. Akartunalı, “Passenger demand fore- casting in scheduled transportation,” European Journal of Operational Research, vol. 286, no. 3, pp. 797–810, 2020

    work page 2020

  2. [2]

    Bright—drift-aware demand predictions for taxi networks,

    A. Saadallah, L. Moreira-Matias, R. Sousa, J. Khiari, E. Jenelius, and J. Gama, “Bright—drift-aware demand predictions for taxi networks,” IEEE Transactions on Knowledge and Data Engineering , vol. 32, no. 2, pp. 234–245, 2018

    work page 2018

  3. [3]

    Deepstcl: A deep spatio-temporal convlstm for travel demand prediction,

    D. Wang, Y . Yang, and S. Ning, “Deepstcl: A deep spatio-temporal convlstm for travel demand prediction,” in 2018 International Joint Conference on Neural Networks (IJCNN) , 2018, pp. 1–8

    work page 2018

  4. [4]

    Particular methods of simulta- neous collection of personal mobility research data from several points,

    A. Tantau, I. Gavrilescu, and L. Fratila, “Particular methods of simulta- neous collection of personal mobility research data from several points,” Energies, vol. 13, no. 22, pp. 6053–, 2020

    work page 2020

  5. [5]

    Integrating household travel survey and social media data to improve the quality of od matrix: A comparative case study,

    Z. Cheng, S. Jian, T. H. Rashidi, M. Maghrebi, and S. T. Waller, “Integrating household travel survey and social media data to improve the quality of od matrix: A comparative case study,” IEEE Transactions on Intelligent Transportation Systems , vol. 21, no. 6, pp. 2628–2636, 2020

    work page 2020

  6. [6]

    Explainable artificial intelligence (xai): Concepts, taxonomies, opportu- nities and challenges toward responsible ai,

    A. B. Arrieta, N. D ´ıaz-Rodr´ıguez, J. Del Ser, A. Bennetot, S. Tabik, A. Barbado, S. Garc ´ıa, S. Gil-L ´opez, D. Molina, R. Benjamins et al. , “Explainable artificial intelligence (xai): Concepts, taxonomies, opportu- nities and challenges toward responsible ai,” Information fusion, vol. 58, pp. 82–115, 2020

    2020

  7. [7]

    Neurosymbolic ai: The 3 rd wave,

    A. d. Garcez and L. C. Lamb, “Neurosymbolic ai: The 3 rd wave,” Artificial Intelligence Review, vol. 56, no. 11, pp. 12 387–12 406, 2023

    2023

  8. [8]

    Gravity model in the korean highway,

    W.-S. Jung, F. Wang, and H. E. Stanley, “Gravity model in the korean highway,” Europhysics Letters, vol. 81, no. 4, p. 48005, 2008

    work page 2008

Show all 27 references
  1. [9]

    A logistic regression model with a hierarchical random error term for analyzing the utilization of public transport,

    C. Wei, T. Lu, and X. Yan, “A logistic regression model with a hierarchical random error term for analyzing the utilization of public transport,” Mathematical Problems in Engineering , vol. 2015, no. 1, p. 430926, 2015

    2015

  2. [10]

    Regional air mobility flight demand modeling in tennessee state,

    K. Acharya, M. Lad, H. Song, and L. Sun, “Regional air mobility flight demand modeling in tennessee state,” in AIAA SCITECH 2025 Forum , 2025, p. 2783

    2025

  3. [11]

    Demand modeling for advanced air mobility,

    K. Acharya, M. Lad, L. Sun, and H. Song, “Demand modeling for advanced air mobility,” in 2024 IEEE International Conference on Big Data (BigData), 2024, pp. 2855–2863

    2024

  4. [12]

    A residual spatio-temporal architecture for travel demand forecasting,

    G. Guo and T. Zhang, “A residual spatio-temporal architecture for travel demand forecasting,” Transportation Research Part C: Emerging Technologies, vol. 115, p. 102639, 2020

    2020

  5. [13]

    Predicting demand for air taxi urban aviation services using machine learning algorithms,

    S. Rajendran, S. Srinivas, and T. Grimshaw, “Predicting demand for air taxi urban aviation services using machine learning algorithms,” Journal of Air Transport Management , vol. 92, p. 102043, 2021

    2021

  6. [14]

    Improving air mobility for pre-disaster planning with neural network accelerated genetic algorithm,

    K. Acharya, A. Velasquez, Y . Liu, D. Liu, L. Sun, and H. Song, “Improving air mobility for pre-disaster planning with neural network accelerated genetic algorithm,” arXiv preprint arXiv:2408.00790 , 2024

    2024 arXiv

  7. [15]

    Modeling travel mode and timing decisions: Com- parison of artificial neural networks and copula-based joint model,

    N. Golshani, R. Shabanpour, S. M. Mahmoudifard, S. Derrible, and A. Mohammadian, “Modeling travel mode and timing decisions: Com- parison of artificial neural networks and copula-based joint model,” Travel Behaviour and Society , vol. 10, pp. 21–32, 2018

    2018

  8. [16]

    Examining nonlinearity in population inflow estimation using big data: An empirical comparison of explainable machine learning models,

    S. Hu, C. Xiong, P. Chen, and P. Schonfeld, “Examining nonlinearity in population inflow estimation using big data: An empirical comparison of explainable machine learning models,” Transportation Research Part A: Policy and Practice , vol. 174, p. 103743, 2023

    2023

  9. [17]

    T. Kim, S. Sharda, X. Zhou, and R. M. Pendyala, “A stepwise inter- pretable machine learning framework using linear regression (lr) and long short-term memory (lstm): City-wide demand-side prediction of yellow taxi and for-hire vehicle (fhv) service,” Transportation Research P...

    2020

  10. [18]

    Analysis of travel mode choice in seoul using an inter- pretable machine learning approach,

    E.-J. Kim, “Analysis of travel mode choice in seoul using an inter- pretable machine learning approach,” Journal of Advanced Transporta- tion, vol. 2021, no. 1, p. 6685004, 2021

    2021

  11. [19]

    Xgboost: extreme gradient boosting,

    T. Chen, “Xgboost: extreme gradient boosting,” R package version 0.4-2, vol. 1, no. 4, 2015

    2015

  12. [20]

    Interpretability of neural networks predictions using accumulated local effects as a model- agnostic method,

    T. Danesh, R. Ouaret, P. Floquet, and S. Negny, “Interpretability of neural networks predictions using accumulated local effects as a model- agnostic method,” in Computer aided chemical engineering . Elsevier, 2022, vol. 51, pp. 1501–1506

    2022

  13. [21]

    Predicting the travel mode choice with interpretable machine learning techniques: A comparative study,

    M. T. Kashifi, A. Jamal, M. S. Kashefi, M. Almoshaogeh, and S. M. Rahman, “Predicting the travel mode choice with interpretable machine learning techniques: A comparative study,” Travel Behaviour and Soci- ety, vol. 29, pp. 279–296, 2022

    2022

  14. [22]

    A unified approach to interpreting model predictions,

    S. Lundberg, “A unified approach to interpreting model predictions,” arXiv preprint arXiv:1705.07874 , 2017

    2017 arXiv

  15. [23]

    Multiscale dynamic human mobility flow dataset in the u.s. during the covid-19 epidemic,

    Y . Kang, S. Gao, Y . Liang, M. Li, and J. Kruse, “Multiscale dynamic human mobility flow dataset in the u.s. during the covid-19 epidemic,” Scientific Data, pp. 1–13, 2020

    2020

  16. [24]

    Population estimates,

    Tennessee State Data Center, “Population estimates,” 2024, accessed: 2024-10-24. [Online]. Available: https://tnsdc.utk.edu/ estimates-and-projections/population-estimates/

    2024

  17. [25]

    Real-time routing with openstreetmap data,

    D. Luxen and C. Vetter, “Real-time routing with openstreetmap data,” in Proceedings of the 19th ACM SIGSPATIAL International Conference on Advances in Geographic Information Systems , ser. GIS ’11. New York, NY , USA: ACM, 2011, pp. 513–516. [Online]. Available: http://doi.acm...

    2011

  18. [26]

    Tennessee economic and fiscal indicators: March 2020 - july 2021,

    Boyd Center, “Tennessee economic and fiscal indicators: March 2020 - july 2021,” 2021, accessed: 2024-10-01. [Online]. Available: https: //haslam.utk.edu/boyd-center/tennessee-economic-and-fiscal-indicators/

    2020

  19. [27]

    County profiles of child well-being in tennessee,

    TCCY, “County profiles of child well-being in tennessee,” 2021, accessed: 2024-10-01. [Online]. Available: https://www.tn.gov/tccy/ data-and-research/county-profiles1.html

    2021

Pith tools

Reviewed August 9, 2026 · model on record in the stance chip above.