论文部分内容阅读
[Abstract]In allusion to how climate change influence regional instability, we collected 14 kinds of representative data about population, climate, production and social structure from 178 countries for the last 10 years. We analysis the indicators related to “the extent that climate change influence regional instability” and the interventions used for “mitigating the risks of climate change and preventing the country from becoming a fragile state” by using BP(Background Propagation)neural network, ISM(Interpretation Structure Model), Three-Factor Analysis Method, STIRPAT Assessment and other methods.
[Keywords]Climate Change; Fragility; Data Mining; BP Neural Network
中圖分类号:R61 文献标识码:A 文章编号:1009-914X(2019)13-0287-02
1、Background
Climate problems loom large in recent decades, the climate change and its adverse effects are issues of common concern for human beings. From the beginning of the Industrial Revolution through now, climate change poses serious challenges to the survival and development of human society. [1]
“Climate change affects every aspect of our life”, former U.S president Barack Hussein Obama Jr. said at a conference in 2015. As one of the world’s most important issues, climate change is not only related to global economic growth prospects and national benefits, but also closely linked to the survival and development of hundreds of millions of people. [2] The destabilizing factors of climate can directly or indirectly affect the fragility of a country. Interacting with poor governance, societal inequalities, and a bad neighborhood, these factors in turn may promote political and economic instability, social fragmentation, migration, and inappropriate responses from governments. [3,4,5,6]Therefore, the climate issue should not be underestimated.
2、Analysis of the Task
We collect data and use BP neural network to solve this problem. We use AdaGrad as optimization algorithm and Sigmoid as excitation function. Specific programs are as follows:
First, we establish the relationship between country’s basic data and country’s fragility, we collected 14 kinds of representative data about population, climate, production and social structure from 178 countries for the last 10 years. In AHP, hierarchy analysis, entropy method, loss function and numerous of methods, we select BP neural network in machine learning. The model is obtained through sample learning. In the process of looking for samples, we collected a lot of data, choosing average annual precipitation and average annual temperature which are related to climate as the input indicators to the model. Then establish the relationship between climate and fragility by using BP neural network. The results show that the larger the output is, the more fragile the country is. 3、The Establishment and Results of the Model
3.1 BP Neural Network
BP (Back Propagation) neural network is a multi-layer feedforward network based on error inverse propagation algorithm. It is one of the most widely used neural network models, composed of input layer, hidden layer and output layer. The network can learn and store a large number of input and output, and does not need to show the mathematical equations of the mapping relationship in advance. The learning law of BP neural network is by using the method of steepest descent, the weight and threshold of the whole network can be adjusted continuously by means of reverse propagation. [8]
3.2 Fragility Assessment
It is impossible to get the absolute truth value, so we based on the data from 178 countries in 12 years [9] and use the Fragile State Index designed by the American peace foundation [10].The table shows that the higher the score, the more fragile the country is. We divide the categories in the scheme above into 3 categories corresponding to the task. Then classify 178 countries according to the fragile state index.Our learning sample is from the data of 35 fragile countries, 89 vulnerable countries and 54 stable countries.
3.3 Process and Results
3.4 Fragility is the cost of changing an interdependent system.
In order to solve the problem, we collected FSI data from 2006 to 2017, data on precipitation and temperature from 2005 to 2014 [11], and data on GDP, CPI, employment rate, population aging, grain reservation, labor force, utilization rate of cultivated land, afforestation, electricity rate and so on. We use these data to establish BP neural network and conduct training and inspection. We obtained about 2000 sets of samples, selected 80% of them for training the network randomly and use the 20% remained for testing to ensure the effect of the model. By comparison, we chose the AdaGrad optimization method. [12]
We establish a network with 12 input layers, 50 hidden layers and 3 output layers.
Firstly, we draw the figure about the error trend of training samples, verification samples, test samples in the training process.The results showed that: through our collection of population, environment, climate, production, social structure, residents’ quality and a series of index data of hundreds of countries, the trained BP neural network after training is highly consistent with the real data, it has a good realistic evaluation significance. Secondly, we draw a figure about the trend of the gradient with the training process, indicating that the smaller the correction is, the more accurate the results are.
Four figures on the left side are scatter plots of real label and prediction label on training samples, verifying samples, testing samples, and all samples. The closer the point is to the line, the more accurate the prediction is.The results show that BP neural network is highly consistent with the real data and has a good realistic evaluation significance.Finally, we draw a figure about the absolute error of the predicts result.
Absolute error is the difference between the predicted value and the real value. It can be seen that most of the error of the sample are near zero, a small number of errors is big because the sample itself are lack of characteristics, this will cause information missing and then misjudge.
By using this model, we judge the national fragility based on the threshold above.
This model describes the relationship between input and output, and establishes an accurate relationship between national vulnerability and input indicators.
Reference
[1]The state council information office of the People’s Republic of China. China’s policies and actions on climate change[N]. The People’s Daily. October 30,2008. (015)
[2]Mo Li. Global leaders “consult” on climate change[N]. The Financial Times. December 3,2015(008)
[3]Ole Magnus Theisen. Is Climate Change a Driver of Armed Conflict? [J]. Climate Change. January 2013. 117(3):613-625
[4]A Picture of the Effects of Global Warming on Countries[N]. Daily Mail. January 16,2015
[5]Wei Zhang, Shaoming Pan, Liguo Cao, XunCai, Kexin Zhang, YihongXu, Wei Xu. Changes in extreme climate events in eastern China during 1960-2013: study of the Huaihe River Basin[J]. Quaternary International. 2015
[6]Haoxian Wang. Study on the Influence of Weather Conditions on Urban Air Pollution[J]. Education of Geography. December 15,2015
[7]China's policies and actions on climate change [J].Progress in climate change research. 2008(06):335
[8]Yun Shi. MATLAB for BP Neural Network[J]. Journal of Xiangnan University. 2010
[9]http://fundforpeace.org/fsi/data/
[10]https://en.wikipedia.org/wiki/Fragile_States_Index
[11]https://crudata.uea.ac.uk/cru/data/hrg/cru_ts_3.23/crucy.1506241137.v3.23/countries/
[12]Waggoner PE, Ausubel JH. A framework for sustainability science: A renovated IPAT identity [J]. PNAS.
作者簡介
曼伯琪,出生年月: 1998.05.10,性别: 女民族:满族籍贯(精确到市):吉林省长春市,当前职务:无,当前职称:无,学历:高中(大学本科在读),研究方向:无。
[Keywords]Climate Change; Fragility; Data Mining; BP Neural Network
中圖分类号:R61 文献标识码:A 文章编号:1009-914X(2019)13-0287-02
1、Background
Climate problems loom large in recent decades, the climate change and its adverse effects are issues of common concern for human beings. From the beginning of the Industrial Revolution through now, climate change poses serious challenges to the survival and development of human society. [1]
“Climate change affects every aspect of our life”, former U.S president Barack Hussein Obama Jr. said at a conference in 2015. As one of the world’s most important issues, climate change is not only related to global economic growth prospects and national benefits, but also closely linked to the survival and development of hundreds of millions of people. [2] The destabilizing factors of climate can directly or indirectly affect the fragility of a country. Interacting with poor governance, societal inequalities, and a bad neighborhood, these factors in turn may promote political and economic instability, social fragmentation, migration, and inappropriate responses from governments. [3,4,5,6]Therefore, the climate issue should not be underestimated.
2、Analysis of the Task
We collect data and use BP neural network to solve this problem. We use AdaGrad as optimization algorithm and Sigmoid as excitation function. Specific programs are as follows:
First, we establish the relationship between country’s basic data and country’s fragility, we collected 14 kinds of representative data about population, climate, production and social structure from 178 countries for the last 10 years. In AHP, hierarchy analysis, entropy method, loss function and numerous of methods, we select BP neural network in machine learning. The model is obtained through sample learning. In the process of looking for samples, we collected a lot of data, choosing average annual precipitation and average annual temperature which are related to climate as the input indicators to the model. Then establish the relationship between climate and fragility by using BP neural network. The results show that the larger the output is, the more fragile the country is. 3、The Establishment and Results of the Model
3.1 BP Neural Network
BP (Back Propagation) neural network is a multi-layer feedforward network based on error inverse propagation algorithm. It is one of the most widely used neural network models, composed of input layer, hidden layer and output layer. The network can learn and store a large number of input and output, and does not need to show the mathematical equations of the mapping relationship in advance. The learning law of BP neural network is by using the method of steepest descent, the weight and threshold of the whole network can be adjusted continuously by means of reverse propagation. [8]
3.2 Fragility Assessment
It is impossible to get the absolute truth value, so we based on the data from 178 countries in 12 years [9] and use the Fragile State Index designed by the American peace foundation [10].The table shows that the higher the score, the more fragile the country is. We divide the categories in the scheme above into 3 categories corresponding to the task. Then classify 178 countries according to the fragile state index.Our learning sample is from the data of 35 fragile countries, 89 vulnerable countries and 54 stable countries.
3.3 Process and Results
3.4 Fragility is the cost of changing an interdependent system.
In order to solve the problem, we collected FSI data from 2006 to 2017, data on precipitation and temperature from 2005 to 2014 [11], and data on GDP, CPI, employment rate, population aging, grain reservation, labor force, utilization rate of cultivated land, afforestation, electricity rate and so on. We use these data to establish BP neural network and conduct training and inspection. We obtained about 2000 sets of samples, selected 80% of them for training the network randomly and use the 20% remained for testing to ensure the effect of the model. By comparison, we chose the AdaGrad optimization method. [12]
We establish a network with 12 input layers, 50 hidden layers and 3 output layers.
Firstly, we draw the figure about the error trend of training samples, verification samples, test samples in the training process.The results showed that: through our collection of population, environment, climate, production, social structure, residents’ quality and a series of index data of hundreds of countries, the trained BP neural network after training is highly consistent with the real data, it has a good realistic evaluation significance. Secondly, we draw a figure about the trend of the gradient with the training process, indicating that the smaller the correction is, the more accurate the results are.
Four figures on the left side are scatter plots of real label and prediction label on training samples, verifying samples, testing samples, and all samples. The closer the point is to the line, the more accurate the prediction is.The results show that BP neural network is highly consistent with the real data and has a good realistic evaluation significance.Finally, we draw a figure about the absolute error of the predicts result.
Absolute error is the difference between the predicted value and the real value. It can be seen that most of the error of the sample are near zero, a small number of errors is big because the sample itself are lack of characteristics, this will cause information missing and then misjudge.
By using this model, we judge the national fragility based on the threshold above.
This model describes the relationship between input and output, and establishes an accurate relationship between national vulnerability and input indicators.
Reference
[1]The state council information office of the People’s Republic of China. China’s policies and actions on climate change[N]. The People’s Daily. October 30,2008. (015)
[2]Mo Li. Global leaders “consult” on climate change[N]. The Financial Times. December 3,2015(008)
[3]Ole Magnus Theisen. Is Climate Change a Driver of Armed Conflict? [J]. Climate Change. January 2013. 117(3):613-625
[4]A Picture of the Effects of Global Warming on Countries[N]. Daily Mail. January 16,2015
[5]Wei Zhang, Shaoming Pan, Liguo Cao, XunCai, Kexin Zhang, YihongXu, Wei Xu. Changes in extreme climate events in eastern China during 1960-2013: study of the Huaihe River Basin[J]. Quaternary International. 2015
[6]Haoxian Wang. Study on the Influence of Weather Conditions on Urban Air Pollution[J]. Education of Geography. December 15,2015
[7]China's policies and actions on climate change [J].Progress in climate change research. 2008(06):335
[8]Yun Shi. MATLAB for BP Neural Network[J]. Journal of Xiangnan University. 2010
[9]http://fundforpeace.org/fsi/data/
[10]https://en.wikipedia.org/wiki/Fragile_States_Index
[11]https://crudata.uea.ac.uk/cru/data/hrg/cru_ts_3.23/crucy.1506241137.v3.23/countries/
[12]Waggoner PE, Ausubel JH. A framework for sustainability science: A renovated IPAT identity [J]. PNAS.
作者簡介
曼伯琪,出生年月: 1998.05.10,性别: 女民族:满族籍贯(精确到市):吉林省长春市,当前职务:无,当前职称:无,学历:高中(大学本科在读),研究方向:无。