RESEARCH ON QUALITY EVALUATION OF LANDSLIDE FORECAST MODEL

A landslide forecast model is mainly used to forecast occurrence time of a landslide. But for model researchers and decisionmakers,an important issue is the model forecast effect. The author proposed a fitting effect index(including posterior index,model fitting efficiency index and RMS error),an experiment forecast effect index(experiment forecast index and related coefficient index),and an establishment quality testing model of landslide forecast. Using gray GM(1, 1)mathematical model,the three exponential smoothing model and time series model fit and experiment forecast the monitoring data of Lianziya dangerous rock body. Prediction results showed that integrated indicators GM(1, 1)is the highest value among the three kinds models. So the landslide using GM(1, 1)to predict is recommended. Through a comprehensive analysis of the model,the author not only analyzes the simulation results,but also a direct comparison fitting results between the models,and provides the adequate basis for the establishment a good model. These show that landslide prediction model for quality testing is an effective and practical method．