Paper: arXiv 2407.13698

Authors: Zijie Pan, Stepan Gordeev, Jiahui Zhao, Ziyi Meng, Caiwen Ding, Sandro Steinbach, Dongjin Song

Abstract

This paper presents a novel methodology for predicting international bilateral trade flows, emphasizing the growing importance of Preferential Trade Agreements (PTAs) in the global trade landscape. Acknowledging the limitations of traditional models like the Gravity Model of Trade, this study introduces a two-stage approach combining explainable machine learning and factorization models. The first stage employs SHAP Explainer for effective variable selection, identifying key provisions in PTAs, while the second stage utilizes Factorization Machine models to analyze the pairwise interaction effects of these provisions on trade flows. By analyzing comprehensive datasets, the paper demonstrates the efficacy of this approach. The findings not only enhance the predictive accuracy of trade flow models but also offer deeper insights into the complex dynamics of international trade, influenced by specific bilateral trade provisions.

Complexity vs Empirical Score

  • Math Complexity: 8.5/10
  • Empirical Rigor: 4.0/10
  • Quadrant: Lab Rats — high math complexity, limited empirical testing

Why this score: The paper features dense mathematical formalizations including Shapley value derivations, Factorization Machine formulations with linear-time complexity proofs, and specialized econometric models (Poisson PML with Lasso penalties), indicating high mathematical complexity. However, empirical rigor is limited as the methodology is presented as a proof-of-concept with described datasets (UN Comtrade, DTAs) but lacks backtesting metrics, implementation details, or live performance validation typical of deployable models.

Research Flowchart

  flowchart TD
  A["Research Goal: Predict Bilateral Trade Flows using PTA Provisions"] --> B["Data: Historical Trade Flows & PTA Text Data"]
  B --> C["Stage 1: Variable Selection via SHAP Explainer"]
  C --> D["Key Provisions Identified"]
  D --> E["Stage 2: Factorization Machine Model"]
  E --> F["Computational Process: Pairwise Interaction Analysis"]
  F --> G["Outcomes: Enhanced Predictive Accuracy & Dynamic Insights"]

Evidence Card

What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card

Proposes

The paper proposes a two‑stage methodology that first uses SHAP explainable machine learning for variable selection of PTA provisions and then applies Factorization Machine models to capture pairwise interaction effects for trade‑flow prediction.

Source sentence
this study introduces a two-stage approach combining explainable machine learning and factorization models.

Assumes 3

  • Country pairs without export information are treated as zero trade flows. data
    Source sentence
    we treated country pairs without export information as zeros and eliminated country pairs with no provision agreements available.
  • Only country pairs that have at least one provision agreement are retained in the analysis. data
    Source sentence
    we treated country pairs without export information as zeros and eliminated country pairs with no provision agreements available.
  • Trade flow values are converted to a binary format (presence vs. absence) for the first modeling stage. execution
    Source sentence
    During the initial phase of our approach, we commence by converting the continuous variables representing trade flows into a binary format, differentiating between the presence (nonzero) and absence (zero) of trade flows.

Data

  • UN Comtrade international trade flows · UN Comtrade macro public
    1968-2017 · all agricultural trade flows
    Source sentence
    This analysis includes data on international trade flows from UN Comtrad, covering all agricultural trade flows between 1968 and 2017.
  • Deep Trade Agreements (DTAs) database · Mattoo et al. text-news public
    1958-2017 · 282 PTAs signed and notified to the WTO
    Source sentence
    We combined the trade flow data with the Deep Trade Agreements (DTAs) database, which was collected by Mattoo et al (Mattoo, Rocha, and Ruta 2020), for the content and evolution of bilateral and regional trade agreements.

Costs

Not applicable The paper reports no trading or execution result.

Finds 2

  • The MLP classifier correctly classifies about 96.09% of training data and achieves 88.29% accuracy and an F1 score of 0.926 on unseen test data for predicting the presence of trade flows. predictive-accuracy
    Source sentence
    Approximately 96.09% of the train data was correctly classified by the model
  • Using the identified 20 key provisions and fixed effects, the Factorization Machine predicts log trade flows with a regression RMSE of 3.26, and the heat plot shows how co‑existence of provisions increases trade flow in log scale. performance
    Source sentence
    The regression RMSE of FM is 3.26 and we display the PTA interaction heat plot as shown in Figure 2. The cell value in the heat plot indicates how much the trade flow will increase in log scale if two provisions coexist.

Checked by others

No other paper in the archive has been found to support or contradict these findings (compared with 16 related papers). Coverage is partial, so this is neither replication nor refutation.

Code

No repository by the authors found in the paper.

Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.

Cited References (36)

  1. Using machine learning methods to capture heterogeneity in free trade agreements (2020)
  2. The breadth of preferential trade agreements and the margins of exports (2022)
  3. Beyond tariff reductions: what extra boost from trade agreement provisions? (2018)
  4. Trade creation and trade diversion in deep agreements (2022)
  5. Do trade agreements stimulate international trade differently? Evidence from 296 trade agreements (2016)
  6. Shaping the world economy; suggestions for an international economic policy (1962)
  7. Handbook of deep trade agreements (2020)
  8. An advanced guide to trade policy analysis: The structural gravity model (2016)
and 28 more
  1. Bias and consistency in three-way gravity models (2021)
  2. Machine learning in international trade research-evaluating the impact of trade agreements (2022)
  3. Factorization machines (2010)
  4. A value for n-person games (1953)
  5. A unified approach to interpreting model predictions (2017)
  6. Accel-gcn: High-performance gpu accelerator design for graph convolution networks (2023)
  7. RRNet: Towards ReLU-Reduced Neural Network for Two-party Computation Based Private Inference (2023) — link
  8. Autorep: Automatic relu replacement for fast private network inference (2023)
  9. MaxK-GNN: Extremely Fast GPU Kernel Design for Accelerating Graph Neural Networks Training (2024)
  10. Advanced language model-driven verilog development: Enhancing power, performance, and area optimization in code synthesis (2023) — link
  11. Empowering Time Series Analysis with Large Language Models: A Survey (2024) — link
  12. Connecting embeddings for knowledge graph entity typing (2020) — link
  13. Connecting embeddings based on multiplex relational graph attention networks for knowledge graph entity typing (2022)
  14. Knowledge graph entity typing via learning connecting embeddings (2020)
  15. National Origin Discrimination in Deep-learning-powered Automated Resume Screening (2023) — link
  16. AQ2PNN: Enabling Two-party Privacy-Preserving Deep Neural Network Inference with Adaptive Quantization (2023)
  17. Learning from Teaching Regularization: Generalizable Correlations Should be Easy to Imitate (2024) — link
  18. APEER: Automatic Prompt Engineering Enhances Large Language Model Reranking (2024) — link
  19. Visual Prompting Upgrades Neural Network Sparsification: A Data-Model Perspective (2023) — link
  20. A length adaptive algorithm-hardware co-design of transformer on fpga through sparse attention and dynamic pipelining (2022)
  21. Time Series Modeling for Heart Rate Prediction: From ARIMA to Transformers (2024) — link
  22. TD3 Based Collision Free Motion Planning for Robot Navigation (2024) — link
  23. Deep Learning Powered Estimate of The Extrinsic Parameters on Unmanned Surface Vehicles (2024) — link
  24. CPSeg: Finer-grained Image Semantic Segmentation via Chain-of-Thought Language Prompting (2024)
  25. Segment Any Building (2023)
  26. Long-Term Visual Simultaneous Localization and Mapping: Using a Bayesian Persistence Filter-Based Global Map Prediction (2023)
  27. PLGSLAM: Progressive Neural Scene Represenation with Local to Global Bundle Adjustment (2024)
  28. A hierarchical spatial transformer for massive point samples in continuous space (2024)

0 of 36 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.