Scholarlink

<p>The economic growth of nations is influenced by the national and international trades especially export trade. The export trade dynamics is influenced by various factors such as global demand fluctuations, exchange rate volatility, trade policies, geopolitical shifts, and domestic production capacities. Volatility is the extent to which market fluctuates and measures the degree of risk. Hence, it is a serious demand to efficiently forecast the India’s export performance. The given study proposes an integrated framework that combines fuzzy logic (FL for human-like, explainable rules) with a neural network (for pattern learning) and LIME ((local interpretable model agnostic explanations) to explain local predictions in a human-understandable form. Our methodology offers integrated adaptive learning power with the linguistic reasoning capability of fuzzy neural network (FNN). The Gaussian membership functions in FL. transforms the numeric input to the linguistic variables (e.g., <i>low</i>,<i> medium</i>,<i> high</i>) that brings tolerance to uncertainty and captures non-linear dependencies among economic parameters affecting trade volumes. Our integrated network learns from the fuzzified input space and understands the inter-relationship between input and output trends by enhancing the transparency using LIME as interpretability module which provides critical decision making where human validation is essential. The proposed approach achieved the highest accuracy of 99.1%, lowest mean absolute error (MAE) of 118.42 and lowest root mean square error (RMSE) of 165.73, thus surpassing conventional models.</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

FNN-LIME: an interpretable fuzzy neural network framework for reliable forecasting of India’s export trade

  • 期刊论文

  • 发表日期：

    2026年1月10日

查看全文
  • Inderjot Kaur,
  • Rajinder Kaur

摘要

The economic growth of nations is influenced by the national and international trades especially export trade. The export trade dynamics is influenced by various factors such as global demand fluctuations, exchange rate volatility, trade policies, geopolitical shifts, and domestic production capacities. Volatility is the extent to which market fluctuates and measures the degree of risk. Hence, it is a serious demand to efficiently forecast the India’s export performance. The given study proposes an integrated framework that combines fuzzy logic (FL for human-like, explainable rules) with a neural network (for pattern learning) and LIME ((local interpretable model agnostic explanations) to explain local predictions in a human-understandable form. Our methodology offers integrated adaptive learning power with the linguistic reasoning capability of fuzzy neural network (FNN). The Gaussian membership functions in FL. transforms the numeric input to the linguistic variables (e.g., low, medium, high) that brings tolerance to uncertainty and captures non-linear dependencies among economic parameters affecting trade volumes. Our integrated network learns from the fuzzified input space and understands the inter-relationship between input and output trends by enhancing the transparency using LIME as interpretability module which provides critical decision making where human validation is essential. The proposed approach achieved the highest accuracy of 99.1%, lowest mean absolute error (MAE) of 118.42 and lowest root mean square error (RMSE) of 165.73, thus surpassing conventional models.