Spatio-temporal Analysis Of National Conflict Event Networks And Conflict Event Detection

International conflict research is an important topic in international relations research,and traditional international relations and conflict research is often qualitative analysis.In recent years,with the maturity of information search and automatic coding technology,large-scale,real-time updated conflict event databases have been established,and quantitative analysis of international relations has become possible,and with the rise of complex network theory and methods in the field of humanities research,it also provides an effective means for quantitative analysis of international relations.However,the reliability of machine-coded conflict databases has been questioned due to coding errors and news bias,and manually coded conflict databases are highly accurate but cannot be updated in real time,resulting in failure to meet the needs of real-time analysis.This paper conducts real-time analysis and detection of conflict events based on the machine-automated coded database,and uses the manually coded conflict event database as a reference for real conflict events,integrates the two databases for quantitative analysis of international relations,and explores the correlation between news conflict events and real conflict events,so as to obtain the identification method of conflict countries and the real-time detection method of conflict events.This paper uses the Global Database of Events,Language and Tone(GDELT)and the Uppsala Conflict Data Program(UCDP)as data sources to construct a national news conflict event network and a national real conflict event network respectively.We use complex network theory to analyze and mine the data to discover the characteristics and evolution patterns of national conflict relationships,analyze the correlation between the networks,establish the connection between the news conflict event network and the real conflict,and finally explore the identification method of conflict countries and the detection method of conflict events based on the discovered correlation characteristics and evolution patterns.The main research work of this paper is as follows.(1)National conflict event network construction and feature analysisThis paper constructs a national news conflict event network based on GDELT data and a national real conflict event network based on UCDP data,and applies complex network theory and methods to analyze them from three aspects: overall characteristics,scale-free characteristics,and node centrality.The results show that the national news conflict event network has the characteristics of small world and scalefree,while the national real conflict events do not have the characteristics of small world and scale-free,indicating that many countries are mentioned in the news of conflict events,but few countries have real conflicts,most of the real conflicts occur within countries,and few conflicts between countries.Moreover,the intensity of conflicts in most countries is small,and only a very small number of countries have a high intensity;there is a correlation between the nodal centrality of countries and the intensity of real conflicts,but there are some differences due to the influence of data saturation and news bias.(2)Spatio-temporal evolution and correlation analysis of the national conflict event networkIn this paper,we construct a time-series national news conflict event network and a national real conflict event network,analyze the time-series changes of the overall statistical characteristics of the networks,the changes of the spatial and temporal distribution of node centrality,and the trends of the network characteristics,and explore the correlation between the networks and between the national news conflict event network and the real conflict intensity,and verify and compare them.The results show that the overall change trends of the two networks are the same,and the national conflict events show a trend of rising and then falling;the spatial distribution of national conflict intensity is not uniform,and countries with higher conflict intensity are concentrated in certain specific regions,such as the Middle East,South Asia,and North America;the change trends of network characteristics can better reflect the changes of real conflicts,and the sudden change of network characteristics often means the outbreak of conflicts,and the conflict There is a big difference between countries with more conflicts and peaceful countries in the degree and range of fluctuation of network characteristics;both the size and trend of network characteristics have high correlation with the real conflict intensity of countries,but the trend of network characteristics has the highest correlation,indicating that the trend of network characteristics better reflects the real conflict intensity of countries.(3)Conflict country identification and conflict event detectionIn this paper,firstly,we select the features with strong correlation with the conflict intensity of countries to classify conflict countries and peace countries,select the optimal classification features according to the classification results,and evaluate the results.Secondly,this paper calculates the relative change number of features and detects the abrupt change of network features using the standard deviation-based anomaly detection method,so as to detect the occurrence of conflict events.The results show that the best classification results can be obtained by using the standard deviation of node strength change of countries as the classification feature for conflict/peace country classification,which indicates that using the standard deviation of node strength change of countries can identify conflict countries better,but the overall classification accuracy is not high,and better classification results can be obtained in extreme cases,while the classification effect is poor in non-extreme cases;using the standard deviation-based anomaly The detection method based on standard deviation effectively detects the mutation of network features and corresponds to the real conflict events,which indicates that the detection of mutation of network features can effectively detect the occurrence of conflict events.