At institutions that organize seasonal concert series, the business problem of retaining existing seasonal subscribers arises. The advertising cost required to retain subscribers that plan on renewing their subscription is significantly lower than to retain the subscribers that do not. Therefore, institutions have incentive to identify ones that fall into the latter category. This thesis presents a comparison of two model variants that allow such predictions. The first one considers a model based entirely on structured data from a data warehouse, while the second one extends that model with non-structured data extracted from online articles using text analysis. Among other attributes, one of the attributes used for concert rankings is the ticket price which varies by concert in a series. Combining the concerts attended by a subscriber during the season, a new attribute is formed which indirectly represents the (dis)satisfaction of the customer with each concert in the series, based on the concert prices determined by the organizer. This thesis offers a novel method for indirect discovery of a customer’s (dis)satisfaction, without subjective evaluation by the organizer. It is based on the sentimental polarity of published articles, their quantity and number of readers. The results of this research have shown a significant divergence between monetary value of individual concerts set by the organizers, when compared to the interest for the concerts shown by the customers. They have also shown that when assessing the subscription renewal probability, the (dis)satisfaction indicator calculated using sentiment analysis of published online content is more relevant than the concert value indicator based on ticket prices.