Research on Social Stability Coefficient based on GWO-XGBoost-LSTM

A “color revolution” achieves regime change by attacking the weakest link in society, but once the ratio of each link in society is out of balance, it will cause social deformation and thus affect social stability. To study the factors affecting social stability and avoid “color revolutions”, it is necessary to establish an effective social stability index system, and on this basis, to build a social stability coefficient model. To establish a complete social stability index system, this paper selects 44 indicators such as the total number of armed personnel, percentage of military funds, total population, per capita GDP, the average age of education, etc. of European and American countries and regions during the period of 1962–2022, and use AHP-EWM model to weight the summation under the premise of dimensionless dimensioning, and then obtain population, defense science and technology, environment, education, economy, and agriculture after dimensionality reduction. The six indicators were then downscaled to obtain population, national defense, science and technology, environment, education, economy, and agriculture, and the social stability coefficient was classified into five levels from 0 to 1 using the TOPSIS evaluation algorithm, with the higher the level the worse the social stability. Using Spearman and Akaike Causality to test the indicator system, it was found that there was a significant correlation and causality between the stability coefficient and each indicator, and the indicator system was well established. To establish a reasonable model of social stability coefficients, this paper optimizes the initial parameters of XGBoost by GWO, fits the prediction to the index system established in Problem 1, then uses CEEMD to process the residuals generated by the GWO-XGBoost model to decompose its nonlinear components, and uses the LSTM model to fit the residuals quadratically, to modify the GWO- XGBoost model. The modified GWO-XGBoost-LSTM model shows superior performance, and its average error is reduced from the initial 0.0218 to 0.0078. For regions that failed to launch “color revolutions
$\prime\prime$
, this paper takes Iran as an example, and after bringing its 44 indicators into the social stability coefficient model in 2018, it comes out that the social stability coefficient of Iran is 0.637, and the social stability level is in the second level of stability to the third level of stability, which is a more stable state at this time. There may be some precursors of social unrest. At the same time, Iran's population development and education level are at a high level, and the general education and knowledge of the Iranian people are high, so the number of people who participate in internal conflicts and revolutions is reduced, which leads to the failure of color revolutions. The article concludes with a summary of this paper's research and submits development proposals in order to prevent color revolutions from posing a threat to social stability.