Application of Ant Colony and Hierarchical Metaheuristic Algorithms in segmentation of Agricultural Knowledge Based Companies

Clustering is one of the most important operations in data mining and its results are useful for researchers and policy makers in various fields for analysis and planning. Since in recent years, the knowledge based economy has been developing with the support of knowledge-based companies in Iran, the analysis of the characteristics of these companies and their segmentation for effective planning provides an appropriate perspective for policy makers.
In this article, Iranian knowledge based companies of agriculture using 2017 data have been clustered based on the field of technology, the number of products and value of product sales. Three clustering methods, simple K-Means and K-Means in combination with hierarchical and ant colony algorithms by using MATLAB 2016a software were applied for clustering. Then, results of three algorithms have been compared and the best one for this data have been selected.
According to the results, the clustering using K-Means in combination with the ant colony algorithm, in comparison with the other two methods, shows a more balanced distribution of the firms among six clusters, and the average Silhouette width value of the 0.7 confirms the validity of this clustering. The highest number of companies and products are in first cluster and the lowest are in 4th and 5th clusters which have the highest average sales per company with 27293 and 5404 million Rials, respectively. The reason for the small number of members in these clusters is that few companies have a large number of knowledge-based products with high sale. In contrast, most companies have acquired a small market share with few number of products. The variety of products and sales of companies indicates the flexibility of the company in different markets, the size of the market and the ability to participate in market development.
Suggestion
Based on the results, it is suggested that similar support programs be designed for companies that are due to similar characteristics in a cluster in order to be effective and avoid losing budget. For knowledge based companies with low average sales that are in a cluster, market support programs and for companies with products diversification and high average incomes, export market development programs are recommended. Providing banking facilities to agricultural knowledge-based companies can not have the same model for every company, and It is necessary to determine the amount of facilities provided and the method of repayment based on the results of the clustering..
JEL Classification: C81, L2, M13, O31, Q16