Scholarlink

<p>A hybrid forecasting framework combining Long Short-Term Memory (LSTM) neural networks with a two-stage least squares (2SLS) system is developed to analyse the global natural rubber (NR) market. Annual data for major producers and consumers (2010–2024) feed a two-layer LSTM with a three-year lookback; forecasts are produced for 2025–2029. Out-of-sample performance is assessed using a 20% hold-out and a rolling-origin backtest, yielding stable forecast errors across folds. Projections indicate that global supply grows from 15.305 Mt (2024) to 17.574 Mt (2029) while demand reaches approximately 16.1 Mt, yielding a supply–demand gap of roughly 1.5 Mt by 2029—a gradually loosening balance rather than a surplus, with supply outpacing demand across all forecast years. NR and synthetic rubber prices co-move, with a soft patch around the middle of the decade followed by firming toward the late 2020s. The geographic composition of supply diversifies as West Africa—especially Côte d’Ivoire—adds capacity, while parts of Southeast Asia, notably Indonesia, contract or stabilising. Structural 2SLS estimates rationalise these paths: demand is primarily income-driven with weak short-run price sensitivity, whereas supply is highly persistent and exhibits a long-run price elasticity close to unity. The combined evidence delivers policy-relevant guidance: prioritise productivity-oriented replanting over area expansion, institute rules-based strategic stocks, upgrade quality and logistics in emerging producer regions, deploy trigger-based trade measures, and strengthen risk-transfer mechanisms for smallholders.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Development of a forecasting model for global rubber demand and supply using LSTM neural networks combined with two-stage least squares (2SLS)

  • 期刊论文

  • 发表日期：

    2026年5月4日

查看全文
  • Montchai Pinitjitsamut

摘要

A hybrid forecasting framework combining Long Short-Term Memory (LSTM) neural networks with a two-stage least squares (2SLS) system is developed to analyse the global natural rubber (NR) market. Annual data for major producers and consumers (2010–2024) feed a two-layer LSTM with a three-year lookback; forecasts are produced for 2025–2029. Out-of-sample performance is assessed using a 20% hold-out and a rolling-origin backtest, yielding stable forecast errors across folds. Projections indicate that global supply grows from 15.305 Mt (2024) to 17.574 Mt (2029) while demand reaches approximately 16.1 Mt, yielding a supply–demand gap of roughly 1.5 Mt by 2029—a gradually loosening balance rather than a surplus, with supply outpacing demand across all forecast years. NR and synthetic rubber prices co-move, with a soft patch around the middle of the decade followed by firming toward the late 2020s. The geographic composition of supply diversifies as West Africa—especially Côte d’Ivoire—adds capacity, while parts of Southeast Asia, notably Indonesia, contract or stabilising. Structural 2SLS estimates rationalise these paths: demand is primarily income-driven with weak short-run price sensitivity, whereas supply is highly persistent and exhibits a long-run price elasticity close to unity. The combined evidence delivers policy-relevant guidance: prioritise productivity-oriented replanting over area expansion, institute rules-based strategic stocks, upgrade quality and logistics in emerging producer regions, deploy trigger-based trade measures, and strengthen risk-transfer mechanisms for smallholders.