FICformer: A Multi-factor Fuzzy Bayesian Imputation Cross-former for Big Data-driven Agricultural Decision Support Systems

Smart agricultural decision support systems leverage big data technology to generate efficient decision recommendations. However, missing values in sensor data can lead to cumulative errors and reduced accuracy in data analysis, compromising the precision of these decision systems. To address this issue, we propose an intelligent multifactor prediction framework for smart agriculture environments. This framework utilizes fuzzy Bayesian data imputation to minimize filling errors and provide a reliable data foundation for decision systems. The framework includes a fuzzy Bayesian module for imputing missing data, resulting in complete datasets and enhancing interpretability. Additionally, a multifactor prediction model is established within an encode-decode framework, incorporating dimension-temporal cross-attention layers to capture and extract correlations among multiple factors. Extensive experiments demonstrate that our proposed model outperforms single-factor prediction, achieving a 13.6% increase in correlation. These results validate the effectiveness of the imputation module in enhancing forecasting precision and reliability, thereby assisting agricultural DSS in meeting evolving demands.

I. Introduction

The widespread application of decision support systems (DSSs) in agriculture and the environment has enabled rapid assessment of agricultural conditions [1]. The rapid development of data-driven DSS provides more favorable and precise intervention decisions in the agricultural production process [2]. In agricultural Big Data-driven systems, accurate environmental forecasting plays a crucial role in providing optimal decision-making [3].