This research endeavors to devise a preprocessing model tailored to handle grain condition data, thereby furnishing criteria for discerning occurrences of mold, condensation, heating, infestation, and similar conditions. Initially, the model assesses the acquired raw grain condition data to pinpoint missing values and anomalies, subsequently opting for interpolation models adaptively based on distinct positions and missing data scenarios. Following this, a blend of statistical and machine learning techniques is employed for feature curation, purging inaccuracies in predictions. Ultimately, a predictive model, such as a supervised learning classification or regression model, is erected to forecast the probability and magnitude of future grain condition incidents. Experimental findings attest to the efficacy of this preprocessing model in managing grain condition data adeptly, thereby heightening the precision in predicting occurrences of mold, condensation, heating, infestation, and related conditions, thereby providing pivotal support for grain safety management.