Research And Application Analysis Of Optimal Reasoning Models Based On Minimum Description Length

At present, with the wide spread of big data and the improvement of data analysis and mining technology, more and more attention is focused on the methods of extracting useful information from big data, which is of a large amount, different kinds, relatively low density value and timeliness. While through using the reasoning models to model and optimize the big data, we can effectively dig out the inherent law or predict the development and change. Therefore, the research of various kinds of reasoning models has always been the focus of data mining research. However, the problem of over fitting is always an important factor to effect the development of reasoning model.In this paper, we mainly study the optimal reasoning model from the given reasoning models. Among them, the optimal reasoning model is the model that solves the problem of over fitting. So the concept of minimum description length is introduced into the reasoning model, and the optimal model is determined by the minimum description length of a series of alternative models which can avoid over fitting and ensure the accuracy of prediction. In the paper, we give a concrete process of computing the number of neurons for an optimal neural network by using the minimum description length method, and analyze the convergence of the algorithm. Furthermore, in order to achieve a successful application, the method is used to predict the traffic conditions and traffic deduction of the floating car data in real time.This paper studies the theory of the minimum description length, and gives a detailed derivation of the minimum description length in the general parametric models. For the neural network reasoning model, due to the error caused by the premature end of the training process in over fitting problem, we give the optimal neural reasoning model based on the minimum description length. Concretely, we study the change of the description length caused by the change of the number of neurons, and choose the minimum description length corresponding to the neurons, namely the optimal neural network reasoning model. At the same time, the paper also studies the properties of the minimum description length in neural network as well as the fitting function in the practical application.Later, this paper uses the improved neural network model to carry on the short-time prediction of the traffic situation based on the floating car data. Mainly, we analyze the traffic situations for the next moment based on the historical floating car data, and verify its accuracy combined with the previous results. Besides, we compare the optimal choice of neural network model under different conditions. Finally, aiming at solving some problems of the existing floating car speed model, an improved speed model is established based on the real-time floating car data, which is used to realize the real-time detection of traffic conditions.