2022年6月11-12日第21届知识与系统科学国际研讨会(The 21st International Symposium on Knowledge and Systems Sciences -- KSS2022)以线上方式举办, 这是KSS会议中断两年后的重启, 得到中国科协2021年度期刊能力提升计划之国际学术交流项目资助而在6月举办. 会议英文投稿经评审录用并由Springer出版了英文论文集(CCIS 1592). 本次会议承办机构在中国, 为发挥本土优势, 鼓励国内学者参加国际会议, 交流最新学术成果, 在第一轮全文投稿结束后, 程序委员会开放英文摘要提交, 满足未赶上第一轮投稿的感兴趣学者的需求. 进一步, 分别开通了一本英文期刊和一本中文期刊的会议全文投稿, 即有英文摘要投稿会议, 同时提交全文相当于直接投稿期刊, 由KSS2022程序委员会评审. 第一轮评审结果反馈作者, 并注册会议作报告, 同时完成论文修改, 会上报告后二轮评审即为终审.
本次专题收录了投稿《系统科学与数学》并在KSS2022报告后最终录用的11篇文章, 它们同属于KSS2022的数据挖掘与文本挖掘的3个分组报告. 这些文章第一作者主要是青年学者, 特别是研究生, 文章的研究全部面向实际问题. 不少文章从网络数据中挖掘有价值的信息, 尤其是从人类各种行为的显性记录基于不同的机理建模而挖掘出有价值的信息充实到相关知识体系中, 展现出对同一个问题求解上数据驱动与传统建模方法的相得益彰, 恰是知识与系统科学国际研讨会所交流的学术成果中突出的特点. 文章的研究来自多种科研项目的支持, 除国家自然科学基金面上项目和学校的一般资助外, 基金重点项目如“自组织视角下的社会群体行为涌现机制和演化规律研究”和国家重点研发计划项目“面向中小微企业的综合质量服务技术研发与应用”起了主要作用. 
本次专题的出版得益于会议与期刊的直接联动. 虽增加了程序委员会的工作量, 但不失为一种工作模式上的创新, 亦有助于青年学者最新研究工作的快速发表. 衷心感谢KSS2022程序委员会参与评审的全体评委(包括在海外高校任教的华人), 感谢期刊主编和编辑部的大力支持与帮助, 期望今后继续这种高效的合作.

The withdrawal of key developers in the open source community will directly threaten the sustainability of open source projects.Therefore,effectively identifying key developers who significantly affect the development work and taking preventive measures can promote the development of community collective wisdom.This paper analyzes Vue and Angular open source projects,focuses on the collaborative behavior of developers,proposes a connection coefficient index to measure the difference of collaborative behavior of developers,and divides developers into three groups and identifies key groups according to the evaluation index system.The results show that the index proposed in this paper is obviously superior to the existing evaluation methods,and the cooperative network suffers greater damage when the cooperative behaviors of the key developer groups detected by this index fail.In addition,key developer types in collaborative networks include nodes with low degree values and non-central locations.This provides a new perspective for further study of cooperative networks.

Currently,most studies on rumor detection mainly focus on social media data and the length of text sequence is short.We argue that existing methods could not capture effective features from health rumors with long texts and then affect the validity of methods.To solve this,we propose an improved BERT-BiLSTM model (I-BERT-BiLSTM),which leverages effective information extracted from texts with long sequences for the health rumor detection.We first conduct text summarization from document-level text.The results are regarded as the input of the deep network model with multi-layer self-attention mechanisms for feature extraction.Finally,we feed the output into BiLSTM for rumor classification.The experimental results show that the model we proposed in this paper achieves 97.75% and 91.15% accuracy on the self-built health rumor data and public data.

In order to identify risk events from Internet media,describe the evolution structures of events and perceive the evolution patterns of event risk,this paper proposes an analysis framework of event risk evolution based on a dynamic network.We construct a time-series network to represent the dynamic development of events,use the Louvain algorithm to identify events,and employ the event transfer metric to construct relation graph between events.Based on the identification of event evolution structure,this paper identifies the main evolutionary paths of events and summarizes the relationship between event risk and event life cycle.The research results show that there are structures of event evolution such as event birth,event merge,and so on.The main paths of event evolution consist of evolution structure.Event risks vary at different stages of the event lifecycle.

Aiming at promoting the high-quality development of the electronic and electrical industry,the status and development trend research of product quality and safety in the electronic and electrical industry is implemented.Selecting the sampling inspection notices issued by the state and 31 provincial and municipal market supervision and administration bureaus from 2018 to 2021 as the data source,this paper applies TextRank algorithm to extract keywords of unqualified items from the published sampling inspection notices of power supply,mobile phone and induction cooker products with a high unqualified rate,and analyzes the key unqualified items;explores the hidden dangers of product safety from three aspects:Overall,time domain and region,and expounds the quality and safety problems of electronic and electrical products in China.The results show that the overall unqualified rate of electronic and electrical products is high,but shows a declining trend,and the coverage of the sampling products is still low,more sampling batches are in southeast coastal regions.Regulatory authorities should strengthen supervision and punishment,carry out the comprehensive sampling inspection of electronic and electrical products;Production enterprises need to strictly control the production process and strictly according to the standard production.

Collecting the food safety supervision and inspection reports issued by the national and provincial market supervision and administration bureaus from 2018 to 2020,this paper analyzes the current situation and evolution trend of China's food industry safety based on entity recognition,text mining and statistical analysis.Firstly,based on BiLSTM-CRF model,the entity names of the inspection category and inspection item in the reports are extracted,and the extracted entity names are combined for rule matching to mine the quality and safety risk information of food industry;Secondly,based on the collected information,this paper studies the causes of high-frequency quality and safety risks in China's food industry with the perspective of time and region,and explores the current situation and evolution trend of food industry safety.The results show that China's food industry is still in a good state with low safety risks;The causes and solutions to part of the quality and safety risks need to grasp the essence of the problems,and the problems exposed in the process of production,storage and sales need to be resolved urgently;Industry supervision needs to be further strengthened,the overall quality of food industry practitioners still has room for improvement,and part of the safety risks should be proceeded from reality.

It has been more than a decade since people got used to engaging in hot social discussions on social media platforms.Fine-grained sentiment analysis of public commons towards social risk events will be helpful for relevant departments to deal with hot issues in time and respond effectively.This paper introduces British sociologist J.M.Barbalet's theory of macrostructural emotion,which affects social structure,order and harmony.And then we construct a related macro-structured emotion lexicon using pre-trained word vector models,combining with psychological human emotion classifications and various emotion lexicons.By taking the"helmet incident"in April 2019 as a case,this paper identifies the main emotion types in Weibo comments under this event,and extracts entities and descriptions highly relevant with the macro structure emotions through point mutual information (PMI) and 2-order dependence distance,so as to obtain the main views and attitudes of the public.

The promotion of new energy vehicles is of positive significance for China to maintain energy security and achieve goals of carbon peaking and carbon neutrality under the new development philosophy.However,fuel vehicles still hold a top post in China's automobile market.New energy vehicles still need to be further promoted,and the influencing factors of new energy vehicle sales remain to be determined.This study uses web crawler technology to obtain online review data about new energy vehicles.It takes 129 new energy vehicles on sale in 2021 as research samples to reveal the internal mechanism of the impact of specific content in online reviews on the promotion of new energy vehicles through text mining and empirical analysis.The study's main conclusions are as follows:1) The number of comments and the emotional polarity of comments will have a positive impact on vehicle sales.2) The contents of "noise control","acceleration performance", "fuel vehicles","battery performance and charging infrastructure"and "weather" in the online reviews of new energy vehicles will have a significant impact on energy vehicle sales.3) The influence of specific content in comments on vehicle sales is heterogeneous in vehicles of different manufacturers and energy types,and the impact of specific content in comments with different degrees of emotional polarity on vehicle sales is also heterogeneous.This study shows that the specific content in online reviews of new energy vehicles will significantly impact new energy vehicle sales.Manufacturers should improve their services according to the user feedback reflected in the related content in the online reviews while conducting online word-of-mouth management.

Aiming at the issues of the high cost of brand value evaluation and the difficulty of collecting consumer comments,this paper proposes a brand value evaluation method for small and medium-sized enterprises by considering online comments.First,based on the framework of the national standard "multi-cycle excess return method for brand evaluation",a brand strength evaluation index system including four dimensions of capital strength,consumers,innovation and social responsibility is built.The indicators of different dimensions are evaluated according to online comments and other data,and the weight of different dimension is determined by AHP method;the brand excess return is calculated by grey forecasting model,and the brand value of small and medium enterprises is finally calculated.Taking 14 small and medium-sized enterprises in food industry as examples,the brand value and the ranking of each enterprise are estimated,and the key factors that affect the ranking of enterprise brand value are explored.The empirical results show that the model has strong applicability to the brand value evaluation of small and medium-sized food enterprises,and also has certain reference value for small and medium-sized enterprises in other industry.

In the process of linguistic decision-making,different decision-makers have different understandings of the same linguistic term,so it is of great significance to consider the personalized individual semantics of decision-makers for the rationality of decision-making results.In this paper,under the environment of linguistic distribution assessments decision matrix and fuzzy preference relation,we construct model to derive personalized individual semantics.A pairwise comparison matrix that satisfies multiplicative consistency is constructed from the comprehensive ranking vector derived from the decision matrix.Then,based on the multiplicative consistency of the fuzzy judgment matrix and the properties of the eigenvectors,starting from the idea of integrating the objective information of the decision matrix and the subjective information of the preference relationship,two methods are proposed to derive the decision maker's personalized individual semantic function:Multiplicative consistency method and eigenvector method.Finally,the individual evaluation information is aggregated to obtain the final ranking of the alternatives.combined the sensory evaluation method,the proposed method are applied to the actual decision-making problem of rice evaluation,and compared with the existing personalized individual semantic derivation methods to show its rationality and validity.

Aiming at the credit risk prediction of small and medium-sized enterprises,this paper proposes a credit risk prediction method based on Long Short-Term Memory (LSTM)-Convolutional Neural Network (CNN) of small and medium-sized enterprises.Firstly,according to the national standard "Enterprise Credit Evaluation Index" and the characteristics of small and medium-sized enterprises,this paper proposes a credit risk prediction index system of small and medium-sized enterprises.The index system includes three kinds of financial and non-financial indicators:Credit intention,credit ability and credit performance.Then,this paper optimizes the network structure and parameters of LSTM-CNN,and applies Dropout and Batch Normalization methods to prevent over fitting.Finally,collecting the information of the listed small and medium-sized enterprises,and after missing value processing,standardization and oversampling,LSTM-CNN is applied to automatically extract features and predict credit risk.The experimental results show that the index system of this paper comprehensively reflect the credit risk situation.The credit risk prediction effect of small and medium-sized enterprises based on LSTM-CNN is better than the comparative models,which overcomes the limitations of traditional methods that cannot dynamically predict the time series data,and ignore the development potential and time continuity of small and medium-sized enterprises.

It is of great significance for regulators,banks and other financial institutions to accurately grasp the credit risk status of listed companies.In this paper,financial indicators and non-financial indicators are integrated to construct a set of credit risk prediction indicators,meanwhile,a combination algorithm SMOTETomekRFE-MLP is proposed for the credit risk prediction of listed companies.The hybrid sampling algorithm SMOTETomek solves the problem of unbalanced sample classification by over-sampling a few samples and under-sampling a majority of samples.Through adding features into the model one by one,recursive feature elimination (RFE) algorithm selects the optimal feature sets based on classification accuracy.Multi-layer Perceptron (MLP) is applied as the binary classifier to predict the credit risk of listed companies.To verify the effectiveness of the algorithm,the base model comparison experiment and ablation experiment are designed to test the algorithm with 3797 A-share listed companies in 2019 as the research object.The results show that SMOTETomek-RFE-MLP credit risk prediction algorithm outperforms the baseline models such as Adaboost,and solves the classification disorder and feature selection problems due to data imbalance,which has certain guiding significance for financial institutions to evaluate the default risk of listed companies.