Datasets:
LittleCxwx
/
paper_data

Dataset card Data Studio Files
xet
 Community
1
text
stringlengths
0
18.7k
"name
”python3”
"language_info":{
"codemirror mode'":{
"name": "ipython
"version": 3
)
"file_extension": " .py ，
"mimetype": "text/x-python，
"name": "python
"nbconvert_exporter": "python'"，
"pygments_lexer": "ipython3"，
"Version ": "3.6.4"
}
)
"nbformat": 4，
"nbformat_minor": 2
第八届泰迪杯数据挖掘挑战赛
“智慧政务"中的文本挖掘应用
摘要
智慧政务即通过“互联网+政务服务"构建智慧型政府，利用云计算、移动物
联网、人工智能、数据挖掘、知识管理等技术，实现由“电子政务"向“智慧政务”
的转变。为了相关部门能够及时、准确地了解到市民的反馈意见和建议并及时解
决群众问题,建立出基于自然语言处理技术(NLP) 的智慧政务系统将具有极大
的推动作用。
针对本次赛题提出的三个问题:
对于第一问, 我们通过对比常见的 6 种文本分类算法, 优先选取了逻辑回归
算法和支持向量机算法进行文本分类, 并建立准确率接近 90%的分类模型， 最后
使用 F-score 评分法进行评价。
对于第二问，我们使用 k-means 聚类算法和 DBSCAN 聚类算法对文本进行
聚类，定义合理的评价热点问题的标准，得到故度前 5 的留言分类。
对于第三问, 我们采用文本相似度对留言的回复情况作相关性的考察, 相似
度是比较两个事物的相似性。本题利用的是余弦相似度, 即将文本映射到向量空
间，再利用余弦距离进行相似度分析。运用相关性的结论以及一定的评价规则，
进一步产生留言回复完整性的 11 类等级划分，综合对留言的回复相关性、完整
性和可解释性这三方面对回复情况进行综合评价。
关键字: 自然语言处理，逻辑回归，支持向量机，k-means 聚类，DBSCAN
聚类;余弦相似度
第八届泰迪杯数据挖掘挑战赛
人Abstract
JIntelligent govemment is to build intelligent govemment through "Intemet 十
govemment service ". using cloud computing. mobile Intemet of things. artificial
intelligence. datamining. knowledgemanagement and other technologies to achieve
the transfomation 位om”e-govemment "to" intelligent govemment " For therelevant
departments to be able to timely and accurately understandthe feedback and
suggestions of the public and solve the mass problems in time. the establishment of a
smart govemment System based on natural language processing technology (NLP)
w刘 have a great role in promoting
Three questions in this competition:
For the first question. by comparing the comimon six text classification
algorithms. we first select the logical regression algorithm andthe support Vector
machine algorithm for text classification. and establish a classification model with an
accturacy rate ofnearly 9096. Finally. weuse theF-score scoring method to evaluate-
For the second question. We Use thek-means clustering algorithm and the
DBSCAN clustering algorithm to cluster the text, define thereasonable criteria for
evaluating hot issues. and get themessage classification ofthe first 3 ofthe heat-
Forthethird question. weuse text simjlarity to examine theresponse ofthe
message. simjlarity is to compare the similarity oftvwo things. This paper uses cosine
simjlarity.that js. text mapping to Vector Space. and then Using cosine 所stance for
simjlarity analysis. Using the conclusions ofrelevance and certain evaluation rules.
the 11 grades ofmessageITesponse integrity are further generated. andthethree
aspects of response relevance. integrity and interpretability are comprehensively
evaluated
及eywords: natural language Pirocessing: logical regresslion: Support Vector machine:
k-means clustering: DBSCAN clustering: cosine similarity
第八届泰迪杯数据挖掘挑战赛