Open AccessClassification of Consumer Goods Safety Cases Based on Improved Bayesian Model
Author(s) -
Yingcheng Xu,
Wei Feng,
Fei Pei,
Haiyan Wang
Publication year - 2020
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1574/1/012157
Subject(s) - weighting , normalization (sociology) , bayesian probability , computer science , multinomial distribution , feature selection , feature (linguistics) , artificial intelligence , class (philosophy) , bayesian inference , data mining , machine learning , pattern recognition (psychology) , econometrics , mathematics , medicine , linguistics , philosophy , sociology , anthropology , radiology
Given the shortcomings of traditional multinomial naive Bayesian model, an improved multinomial Bayesian model based on feature weighting is first proposed. Second, in the weight setting of the model’s feature selection items, the TF-IDF algorithm does not consider the distribution of feature items between and within classes, and a range of factors related to word distribution is proposed such as intra-class factors, inter-class factors, document set factors, and normalization factors, then an improved TF-IDF weighting algorithm is put forward. Finally, combined with the case of consumer product safety, the effectiveness and feasibility of the proposed model and algorithm are verified.