Font Size: a A A

Construction And Application Of Weighted Bayesian Model

Posted on:2022-12-10Degree:MasterType:Thesis
Country:ChinaCandidate:C X HongFull Text:PDF
GTID:2518306743474424Subject:Computer technology
Abstract/Summary:PDF Full Text Request
Naive bayes(NB)algorithm is a classification algorithm based on bayes principle,based on probability knowledge,and has a solid mathematical foundation.It has the advantages of easy construction,easy understanding and high efficiency,it has always been one of the most commonly used algorithms in machine learning,and has been widely used in the field of face recognition.However,the strong classification performance of the NB algorithm is based on a conditional independence assumption that hardly holds in reality,for data with strong dependencies,the classification accuracy will be greatly affected.To solve this problem,many improvement methods have been proposed,among which the attribute weighting method has achieved good results.However,how to weight the attributes and obtain the best combination of weights is still a very difficult problem.In response to this problem,this paper proposes several improved models of the NB algorithm,the main innovations of the paper are as follows:1.Attribute weighted naive bayes classification model based on a space search optimization algorithm is proposed.The model combines the attribute weighting method with space search optimization algorithm,by adding weights to the basic NB calculation,uses a one-dimensional vector to represent the relationship between different attributes,and optimizes the weight vector based on the space search optimization algorithm,through multiple iterations,a set of optimal weight vectors is found,so that the model classification effect is optimal.The model is tested on the UCI standard dataset,and the results show,compared with the traditional NB algorithm and some recent advanced models,SSOA-WNB has better classification accuracy.2.A class-specific attribute weighted bayes model based on optimization algorithm is proposed.The weights are treated with a more fine-grained attribute weighting approach,where the weights are also treated in the vertical axis,that is,different weight values are assigned to the same attribute in different categories,and the weight matrix is a two-dimensional matrix.We constructed two classification models based on optimization algorithms.The first one,a class-specific attribute weighted naive bayes model based on the particle swarm algorithm is constructed,and obtain the optimal two-dimensional weight matrix based on particle swarm optimization.The second,a class-specific attribute weighted naive bayes model based on a space search optimization algorithm is constructed,and uses the space search optimization algorithm to optimize the two-dimensional weight matrix.3.A class-specific attribute weighted bayes application model based on optimization algorithm is proposed.Due to the high classification performance of the class-specific weighted naive bayes model based on a space search optimization algorithm,the model is applied to face recognition.First,the face samples are preprocessed,and each image is turned into one-dimensional vector data.Then,the principal component analysis method is used to reduce the dimensionality of the data,and the feature values are extracted.And finally classified using class-specific weighted bayes model based on space search optimization.The experiment has achieved a classification effect superior to the traditional model.
Keywords/Search Tags:Naive bayes, Attribute weighting, Particle swarm optimization, Space search optimization algorithm, Face recognition
PDF Full Text Request
Related items