Quantitative Prediction Model Of Survival Time Based On Clinical Follow - Up Data

Survival time is defined as the duration that spans the beginning till ending of events. Taking account of survival time as a research target, survival analysis is referred as a scheme which builds the model for survival time prediction or factor analyzing. The survival time analysis is a wildly used method applied for clinic medicine, economy, engineering and so on, where statistics and data mining are two basic means for modeling of the survival time. However, the difficulty of the modeling is that the training set contains censored data samples.There are two pivotal ways to deal with the problem of censored data samples: the first way is to drop the samples during the training process of modeling; and the second way utilizes those samples based on rational hypothesis. Usually for the censored data, the recorded time is used as the lower bound for survival time.This paper studies the typical survival time prediction statistical model and data mining model, and proposes two schemes to quantitatively predict survival time based on data mining technology.The first scheme of proposal is based on the hypothesis of the lower bound for the survival time. Via the results of experiment comparison, it analyzes the advantages of understandability. However, there exists shortages of the performance verification of censored sample processing. And it is required to propose a model by new hypothesis for performance improvement.The second method uses new assumptions for censored data samples processing based on distance metric characteristics. According to this, this paper presents a model which is based on secondary learning style for survival time modeling. In this part, the proposed scheme illustrates the form of survival time predication model, the training algorithm and deducing the time complexity of training algorithm. Survival curves of the comparative experiment and accuracy of predication comparative experiment depict the performance of proposed quantitative forecasting model, which outperforms other models based on data mining in the collected data set. And this model can hence be used as a novel scheme for clinical quantitative prediction of survival time.