In highly competitive market, besides competitive price and quality, providing proper client support is a crucial factor in increasing customer satisfaction. Since customer support provision mostly involves human interaction, selecting a suitable agent that not only has necessary knowledge to handle a customer request but is also a proper match with the customer from personality point of view can highly increase the probability of making the client satisfied with the support service. This paper exploits machine learning techniques to train a model using history of the results of previous interactions for selecting a suitable agent for a customer request based on the extended profiles of the agents and the customer. Data analysis and machine learning algorithms have been applied to extend the profiles of the agents and customers to help the model in selecting the best match based on the probability of success for an interaction. The experimental results on the historical data of a large call center in Portugal show the proposed model can improve the interaction success rate.