A binary classification method based on seepage analysis

Abstract

The invention provides a binary classification method based on seepage analysis, which comprises the following steps: 1, constructing an index network based on a data vector; 2, seepage analysis of index network; 3, constructing a likelihood function and determining a key threshold; 4. evaluation and validation of the model. Through the above steps, the invention further provides technical supportand theoretical support for large-scale, high-dimensional and high-complexity system group classification and evaluation classification effect based on seepage analysis; in addition, the background knowledge is expressed by graph model and the network analysis is carried out by using the seepage theory based on phase change, which reduces the computational complexity, converges quickly and is suitable for large-scale computation and reduces the computational cost.

Description

technical field [0001] The present invention proposes a binary classification method based on seepage analysis, constructs an index network based on the degree of correlation between individual units, constructs a binary classifier through network seepage analysis, and uses a confusion matrix for model evaluation, which belongs to machine learning and network Interdisciplinary fields of science. Background technique [0002] Binary classification problems are widely used in fields such as medicine, industry, and social analysis. With the advent of the era of big data, data has become an indispensable part of people's lives, forming a way of life from data to data; the emergence of mobile Internet has largely enriched the way people generate data At the same time, with the emergence of modern theories and technologies such as artificial intelligence, machine learning and cloud computing, it provides a strong guarantee for the solution of classification problems, such as: log...

AI Technical Summary

Problems solved by technology

In 1958, the famous British statistician David Cox discussed the logistic regression model, which is mainly applied to classification problems with linear features. The model has limitations in adapting to data and scenarios. When the dimension of the feature space is large, The performance of logistic regression is poor, and it cannot handle a large number of multi-category features or variables well; in addition, in 1963, VladimirN. In constructing hyperplane or hyperplane set, it is used for classification, regression or other tasks, such as: anomaly detection, etc., but the SVM algorithm is difficult to implement for large-scale training samples, because SVM uses quadratic programming to solve support vectors, and solving two The second plan will involve the calculation of the m-order matrix (m is the number of samples). When the number of m is large, the storage and calculation of the matrix will consume a lot of machine memory and computing time; K-means clustering (k-means) initially Proposed by Stuart Lloyd in 1957, as a pulse code modulation technology, with the development of information technology, it is widely used in data clustering analysis. This algorithm needs to continuously adjust the sample classification and continuously calculate the adjusted new cluster. Class center, so when the amount of data is very large, the time overhead of the algorithm is very large; K-neighborhood algorithm (KNN) is mainly applied to data mining and classification. This algorithm is a lazy algorithm. When the amount of calculation for classification is large, it is necessary to scan all The calculation distance of training samples has a large memory overhead and slow scoring speed, resulting in increased calculation costs

Images