Research On Prediction Methods Based On Credibility In Big Data

Data-based prediction is a research content that is widely concerned in academic and application.Data mining methods proposed by many scholars have solved the problem of prediction on some extent,but the systematic method of processing big data in theory is still not perfect.Therefore,this article has done the following two tasks in response to the large amount of data in big data:1)In big data with structural characteristics,a classification rule acquisition method based on decision tree by the combination of sampling with and without replacement(SDTI)is studied.The research includes the screening mechanism of initial classification rule base,the refining strategy of classification rule base and the method for determining the credibility of the classification rules based on sampling with replacement.Combined with the theoretical and experimental results,the feasibility and interpretability of the method are analyzed which provides theoretical and methodological support for the classification rule acquisition method in big data.2)In big data with structural characteristics,the uncertainty prediction method based on credibility(RDP)is studied.The research includes the implementation mechanism of RDP and the method of determining credibility.The approximation of the decision attribute values distribution is analyzed by the combination with the law of large numbers and the experimental results,which shows that RDP has good interpretability and operability,and can provide theoretical and methodological support for prediction in big data.Therefore,the models proposed in this paper all have good interpretability and structural characteristics,which can help managers make reasonable decisions in uncertain environments.