For the purpose of the exercice, I've to analyse international's chicken market. Such analyzes will help us firstly getting a better understanding of global chicken trade market and on the other hand finding out new trade agreement over the world.
How it works ?
Before bringing the countries together, we have to decide what are the best features to match with our vision. Political, Economical, Social, Technological, Ecological and Legal are good topic to begin.
For example, Political criteria bring us more information about government's stability (No conflict).
As GDP's growth feature give us some usefull information about the economic health and development of a country.
Go deeper on Pestel analyze with a full detailed application available here
I enjoy open world datawarehouse from World Bank Group and Food and Agriculture Organization to build my entire training set. I highly recommend to checkout their website as I did.
- Every data exported from WBG and FAO are available in Data Source repository
- Complete training set as a results of step by step data preparation program that I used to feed my clustering model
PCA results and interpretation are fully exposed in Clustering notebook.
I try two differents clustering approach with hierarchical clustering firstly following by the famous Kmeans. All of those 2 methods are based on distance calculation between each point and the nearest or further centroid.
NB : Hierarchical clustering is set up with longest point-cluster distance while the reverse is operated for Kmeans.
For the features I used to feed the model, take a look into acp_training_set
As a result of our clustering, I identified 2 interesting clusters :
- Cluster with underdeveloped countries struggling with social inequalities and food deficit
- Cluster with developed countries, stable government and low international chicken trade