Food testing data is an important tool for food risk analysis. The final data matrix is missing due to different testing items for similar foods， and most of the existing food testing data is undetected. Through the introduction of TF-IDF (The term frequency-inverse document frequency) weight determination method has constructed a new type of food risk analysis model. This paper uses the sampling information of vegetable samples of edible agricultural products in a city from 2019 to 2020 as the research data， and calculates the risk index of each sample in the vegetable through the model. The results show that among the vegetable products tested from 2019 to 2020， the high-risk index is leeks and celery， and the over-standard index is chlorpyrifos， which needs to be paid more attention in supervision， while most of the remaining vegetables are low-risk. Compared with other traditional analysis methods， this analysis model can give a specific risk index， is intuitive in evaluation， and shows better evaluation performance in big data analysis. At the same time， this model sets weights in an objective and universal mode， which eliminates the influence of subjective factors in the evaluation and further enhances the practicability in diversified data analysis. The model is established in the context of the era of big data， and provides a new way of thinking for further in-depth research and exploration of new paths for food safety risks and evaluation methods.