In marketing, especially in targeted marketing, one of the most frequent problems is dichotomous in nature. For instance, to create an advertisement or to know which good to promote, the company needs to have such information as whether the customer prefers item A or prefers item B, buys or does not buy a product, is likely to churn or not. The solution to these problems gives opportunities for companies to create new strategies, increase their revenue and save the most precious thing – time, by automating all of the processes. Nowadays, mentioned problems can be resolved by supervised machine learning, to be more precise, by its division - binary classification. To use a binary classification model, a company needs to collect data about their products, customers, transactions, and other relative information concerning the sales. Moreover, a historical prospect and customer data of the company should be divided into two necessary groups, like a buyer and non-buyer for instance. During the model training, the classifier learns the positive attributes of firm’s most desirable buyers and the negative attributes of its non-perfect buyers. Thus, allowing the finalized model to be used to score the existing database, and in addition, new prospects in the future for their possibility of becoming a buyer. This thesis focuses on the explanation and implementation of different binary classifications models, constructed to solve one of the cases described above, to be more exact, to build a model which will identify potential future buyers of TV. The dataset for this problem is from a fictional company called Pentatonic which specializes in selling high-end televisions.