Ridge regression numerical prediction method based on Markov chain

Abstract

The invention relates to a ridge regression numerical prediction method based on a Markov chain, and belongs to the field of machine learning and data mining. The method is characterized by comprisingthe following steps: (1) determining the output of a model; (2) determining a loss function L (l); (3) determining the output of the model according to the formula (1), (3) determining a value; (4) determining a residual sequence E = {e1, e2, and the like, en}; (5) determining an m-step state transition matrix p (m); (6) establishing a Markov chain to correct the prediction result according to the known state transition matrix p (m) and the initial state ei; and (7) taking the test set samples as input, and carrying out model training to obtain a prediction result. According to the ridge regression numerical value prediction method based on the Markov chain, firstly, the ridge regression model is established for prediction, then the Markov chain is used for correcting the residual error generated by prediction, and numerical value accurate prediction is achieved. According to multiple groups of data experiment results, compared with other models, the prediction method for enhancing the generalization ability of the model on the basis of ensuring the prediction precision is provided for numerical prediction.

Description

technical field [0001] The invention relates to the fields of machine learning and data mining, and mainly relates to a numerical prediction method. Background technique [0002] At present, for numerical prediction problems, most models can achieve high prediction accuracy, but the models are prone to overfitting. With the improvement of training ability, the prediction ability decreases instead. These models often show good predictive performance on the training set, but on the test set or some unknown data, the predictive ability will be greatly reduced. Especially the neural network, although the neural network can achieve high prediction accuracy in numerical prediction, it is prone to overfitting, and the model is too complex and the training time is too long. Taking BP neural network as an example, the early BP neural network has a series of problems such as weak approximation ability, slow convergence speed, and easy to fall into local optimum. The generalization a...

AI Technical Summary

Problems solved by technology

The network tends to fall into a local minimum point during the training process. Although a series of parameter optimization methods have been proposed, such as genetic algorithm, cuckoo search, and grid parameter optimization, etc., these methods have complicated calculation processes and a large amount of calculation. The algorithm needs to pass through N generations of reproduction and survival of the fittest through multiple crossover and mutation methods, and it takes a lot of time to adjust parameters

greatly reduces the applicability of the model

Images