A three-layer joint dynamic selection optimal model method for large data volume forecasting

Abstract

The invention provides a large data volume prediction three-layer combined dynamic selection optimal model method including three layers of a prediction model algorithm library, a weight algorithm library and an optimal weight algorithm selection algorithm. The prediction model algorithm library is arranged at the bottommost layer. The weight algorithm library is arranged above the prediction algorithm model library, and the optimal weight algorithm selection algorithm is arranged above the weight algorithm library. According to the large data volume prediction three-layer combined dynamic selection optimal model method, the three-layer structure has four characteristics of high expansibility, prediction stability, the dynamic adjustment characteristic of the model and no distinctiveness of the prediction data for the model. The combined algorithms are applied in the method, certain disadvantages of the common algorithms are avoided by the algorithm, multiple algorithms are organically combined by using the method of assigning weight to multiple models, high weight is assigned to the most adaptive algorithm and low weight is assigned to the relatively poor algorithm so that the accuracy of data prediction can be guaranteed and the prediction stability after increasing of data length can also be guaranteed.

Description

technical field [0001] The invention relates to a three-layer joint dynamic selection optimal model method for large data volume prediction. Background technique [0002] Up to 250 terabytes of data are now generated every day, more than 90% of the total data generated in the past two years. A large amount of data is stored in a computer in a structured form. After these data are structured, while being convenient for storage, they also lose their logical connection. For example, two adjacent communities in communication affect each other, cause and effect each other, and continue into the future in a certain pattern, while What is stored in the computer is only two columns of data, without association and pattern. In reality, there may be countless columns of such data, which makes the association and pattern hidden deeper and the form more complicated. In such a large and complex data, to discover associations, capture patterns, and predict the future requires a stable a...

AI Technical Summary

Problems solved by technology

With the accumulation of data or when predicting long-term data, the problem will become prominent, and the prediction effect will become poor

[0005] The second problem is that when a large amount of different data is to be predicted, a model needs to be selected for each column of data. In this way, it takes a lot of time to select the model. Even if this is done, the above-mentioned problem cannot be avoided - model selection error , and hope that the selection process of each model is simple and scientific, and the model prediction results are stable and relatively accurate

[0006] The third problem is that fast dynamic prediction cannot be achieved

Obviously, this is not enough for fast and dynamic forecasting

Images