Applied Research Of Uncertain Bayes Algorithm In Prediction Of Landslide Hazard

With the development and progress of science and technology, the computer technology has been gradually integrated into the various fields. The Bias classification technology applied to the loess landslide prediction research in the field is more and more concerned by scholars. The Naive Bias classification technology is the traditional batch learning mode. All the accuracy data need to be ready before the model studying. However, the rainfall which leads to loess landslides is a interval uncertain data and as time changes, the past data will change in time and the new data is increasing. Therefore, it is difficult to acquire effectively the required parameters during the establishment process of constructing the landslide hazard classification and prediction model and to express the quantitative characterization of landslide inducing factors. In order to better represent the uncertain data, you’ll need to use the uncertain data mining technology on uncertain data for quantitative analysis. In order to make the classification model better adapt to the new data or changed data, we urgently need to research on one method of incremental learning, incremental update knowledge, and to revise and improve the previous knowledge, make the updated knowledge to adapt to the newly added data.First of all, to effectively describe the attributes of rainfall which is an attribute level uncertainty and a difficult process of constructing quantitative classification model characterizing the uncertainties of the interval attribute data, this article construct a Uncertain Bias classification model by introducing the Uncertain Bias algorithm and determining the possible world model with the naive Bias classification model data together.Secondly, aiming at the difficulty to obtain the complete data to construct the prediction model and when updating or increasing the result parameters make the existing Bias classification method of effective prediction accuracy low, we put forward the uncertain Bias incremental algorithm on the base of the incremental learning and Bias classification. The incremental bias model gets the useful information from the new data, and automatic updates the classification model by self. It has the lower dependence of the whole training tuples and higher recognition rates for the new added data.Then, this paper summarizes the prediction model of the whole classification engine, carries on the detailed introduction to the classification engine of each sub module, describes the process of building the data warehouse in detailed, and illustrates the process dimension, data cleaning, data sheets and projection transform combined with cuts in data preprocessing.Finally, this article designs a series of contrast test between the Uncertain Bias classification model with the naive Bias classification model and the Uncertain Bias incremental classification model with naive Bias classification model and the Uncertain Bias classification model on the foundation of the loess landslide points which are determined and uncertainty feature set. The experimental compare the model from the aspects of the effective prediction accuracy, the overall prediction accuracy and time requirements. The experimental results show that, in the case of complete data, the Uncertain Bias algorithm effectively represent the uncertain interval data, and its effective classification accuracy and the overall accuracy are better than the naive Bias classification algorithm; when data are incomplete, the uncertain Bias incremental classification algorithm has better robustness, accuracy, that make up the shortcomings that the traditional Bias classifier and uncertain Bias classifier are rebuilt which spend a lot of time and resources.