Abstract: Because there is a large error defect of predicting the severe fluctuating fire data by traditional grey forecasting model, the article expends the original data sequence of grey prediction model using Taylor formula and expends Lagrange type remainder, which modified the traditional grey forecasting model and improved the traditional grey prediction algorithm. The article used Matlab software to program, and used the traditional gray prediction algorithm and improved gray prediction algorithm,respectively.Then the article selected three sets of data to predict aiming at the mini zone bathtub shape fire accident. As the results show that: modified grey forecasting model prediction curve gives satisfactory accuracy comparing to the characteristic of bathtub curve, which has a high consistency with the original data.And the poor quality, large error and low accuracy is improved. Compared with the traditional grey prediction model, the error of the improved grey prediction model has reduced 86.59%, 55.32%, and 55.70%, respectively. The prediction results of the improved grey prediction model can meet the requirements in three aspects: accuracy, relative error and correlation.

Key words: Fire accident, Prediction, Bathtub shape, Gray, Matlab

摘要： 基于传统灰色预测模型预测剧烈波动型火灾数据误差较大的缺陷，使用泰勒公式将灰色预测模型的原始数据序列展开，并展开拉格朗日型余项，以此修正传统的灰色预测模型，改进了传统的灰色预测算法。使用Matlab软件编程，分别采用传统的灰色预测算法和改进的灰色预测算法，针对小区间浴盆形火灾事故，选取三组数据进行预测。结果表明,修正后的灰色预测模型改善了传统灰色预测模型预测浴盆形火灾事故稳定性差、误差大、精度低的不足，预测曲线符合浴盆形曲线特征，与原始数据曲线特征具有较高的一致性。改进的灰色预测模型与传统的灰色预测结果相比，三组数据的平均相对误差分别降低86.59%、55.32%、55.70%。改进的灰色预测模型预测结果在精度、相对误差、关联度三方面都能满足要求。

关键词: 火灾事故, 预测, 浴盆形, 灰色, Matlab