«上一篇/Previous Article|本期目录/Table of Contents|下一篇/Next Article»

《计算机技术与发展》[ISSN:1006-6977/CN:61-1281/TN]

卷:

30

期数:

2020年09期

页码:

109-114

栏目:

安全与防范

出版日期:

2020-09-10

文章信息/Info

Title:

Research on Food Safety Early Warning of Multi-label Classification of High Dimensional Data

文章编号:

1673-629X(2020)09-0109-06

作者:

甄俊涛; 刘 臣

上海理工大学 管理学院,上海 200093

Author(s):

ZHEN Jun-tao; LIU Chen

School of Business,University of Shanghai for Science and Technology,Shanghai 200093,China

关键词:

高维数据; 数据降维; 最大依赖性降维方法; 长短期记忆神经网络; 多标签分类

Keywords:

high dimensional data; data dimensionality reduction; maximum dependence dimension reduction method; long and short term memory neural network; multi-label classification

分类号:

TP183

DOI:

10. 3969 / j. issn. 1673-629X. 2020. 09. 020

摘要:

随着大数据、互联网等新兴科技的飞速发展,人们生活逐步向数字化、信息化迈进,高维图像数据、高维文本数据等各类复杂数据不断涌现。 高维数据具有包含信息量大、易出现信息冗余的特征,给文本分类带来阻碍。 为此,提出一种基于长短期记忆神经网络(LSTM)的高维数据多标签分类方法。 该方法从数据降维的角度出发,利用最大依赖性降维方法(MDDM),将高维数据降为低维数据,提高有效信息占比、减少信息冗余。 将降维后的低维数据作为长短期记忆神经网络的输入,利用 softmax 函数对神经网络的输出进行多标签分类。 在食品稽查数据上进行的安全预警实验验证了该方法的可行性,最终分类准确率达到 86.5% ,比未降维的数据分类准确率提高 36.5% 。 实验还对比了不同神经网络模型在该数据集上的分类性能,结果表明使用 LSTM 神经网络进行分类结果较好。 良好的分类结果表明该方法在食品稽查数据集上特征提取的准确性,食品安全稽查部门可对具有该违法特征的食品生产企业进行监督管理,从而避免食品安全问题的发生,以达到食品安全预警的目的。

Abstract:

With the rapid development of big data,Internet and other emerging technologies,people爷 s life is gradually moving towards digitalization and informatization,and high-dimensional image data, high-dimensional text data and other complex data are emerging,which contains large amount of information and is prone to information redundancy,and also hinders text classification. For this,we propose a multi-label classification method for high-dimensional data based on long and short term memory neural network (LSTM).From the perspective of data dimensionality reduction,the maximum dependence dimensionality reduction method (MDDM) is used to reduce the high-dimensional data to the low-dimensional data,improve the proportion of effective information and reduce information redundancy. Then the low-dimensional data is taken as the input of LSTM,and softmax function is used to classify the output of the neural network. Experiments of early warning on food inspection data is to verify the feasibility of the method, and the final classification accuracy reaches 86.5% ,which is 36.5% higher than that of data without dimensionality reduction. The experiment also compares the classification performance of different neural network models on the data set,and the it shows that the LSTM neural network has better classification results which indicate that the proposed method is accurate in feature extraction from the food inspection data set,and the food safety inspection department can supervise and manage the food production enterprises with the illegal characteristics,so as to avoid the occurrence of food safety problems and achieve the purpose of food safety warning.

常用功能

更新日期/Last Update: 2020-09-10