In Eastern India, a widespread practice known as “rice fallow pulse” (RFP) involves using the soil’s remaining moisture to grow a short-duration pulse crop. For rainfed systems, it is an excellent practice of climate adaptation. To help farmers make informed decisions about where to plant what and to help policymakers create favorable conditions for timely seed distribution, it is imperative to forecast the appropriateness of pulse crops both geographically and temporally. Using fuzzy AHP (FAHP)-based machine learning methods, we tried to detect pulse appropriateness both geographically and temporally while considering fifteen natural, climatic, environment, and soil health-related characteristics in the Western Lateritic Zone of the Indian State of West Bengal. According to the findings, all machine learning (ML) techniques identified high-suitability zones in the districts of Murshidabad, Birbhum, Paschim Bardhaman, Paschim Medinipur, and Jhargram. By using machine learning techniques such as shrinkage discriminant analysis (SDA), neural network (nnet), random forest (RF), Naive Bayes (NB), rule-based C5.0, genetic algorithm (GA), and particle swarm optimization (PSO), it was found that moderate suitability zones were visible in some areas of Murshidabad, Birbhum, Paschim Bardhaman, Paschim Medinipur, and Purulia. Additionally, it was noted that all ML approaches revealed maximum low suitability zones in certain areas of Birbhum, Bankura, Purba Bardhaman, Purulia, and Murshidabad. Finally, district-level yearly pulse yields of minor, chickpea, and pigeonpea verified the precision of the ML-based models. We have devised a structure to assess pulse suitability analysis to improve crop and land productivity. One of the world’s most populous regions can use the data to inform policy decisions that will improve food and nutritional security in the face of shifting economic and environmental conditions.

在印度东部，一种被称为“水稻休耕豆类”(RFP) 的广泛做法涉及利用土壤剩余的水分来种植短期豆类作物。对于雨养系统来说，这是适应气候的绝佳实践。为了帮助农民就在哪里种植什么做出明智的决定，并帮助政策制定者为及时分发种子创造有利条件，必须预测豆类作物在地理和时间上的适宜性。使用基于模糊层次分析法 (FAHP) 的机器学习方法，我们尝试在地理和时间上检测脉冲适当性，同时考虑印度西孟加拉邦西部红土带的 15 个自然、气候、环境和土壤健康相关特征。根据调查结果，所有机器学习 (ML) 技术都在 Murshidabad、Birbhum、Paschim Bardhaman、Paschim Medinipur 和 Jhargram 地区确定了高度适宜性区域。通过使用收缩判别分析（SDA）、神经网络（nnet）、随机森林（RF）、朴素贝叶斯（NB）、基于规则的C5.0、遗传算法（GA）和粒子群优化等机器学习技术（ PSO），发现在 Murshidabad、Birbhum、Paschim Bardhaman、Paschim Medinipur 和 Purulia 的一些地区可见中等适宜区。此外，值得注意的是，所有 ML 方法都揭示了 Birbhum、Bankura、Purba Bardhaman、Purulia 和 Murshidabad 某些地区的最大低适宜区。最后，区级小豆、鹰嘴豆和木豆的年产量验证了基于机器学习的模型的精度。我们设计了一种评估豆类适宜性分析的结构，以提高作物和土地生产力。世界上人口最多的地区之一可以利用这些数据为政策决策提供信息，从而在经济和环境条件不断变化的情况下改善粮食和营养安全。