Improving of the Effectiveness of Television Advertising Based on the Analysis of Data on Television Viewing Using Machine Learning Methods and Neural Networks

The theme of the master's project - " Improving the effectiveness of television advertising based on the analysis of data on television viewing using machine learning methods and neural networks" - which considered the trends in the development of TV advertising in Russia and the world, the general problems in making predictions on TV viewing, the distribution of television and as a consequence, the need to use artificial intelligence technology to work with the data, their practical use in this project and the formalization of the results obtained. The basis for analysis in the project work is the initial data of the ratings for television viewing for 2017 from the research company Mediascope. The aim is to identify the impact of selected features used in the construction of predictive models on TVR by audience groups, in order to effectively use advertising budgets. The research methods used in the project are subdivided into: initial analysis of the received data, preparation of data for their further processing and construction of models using machine learning, neural networks and methods of processing of natural language for construction of models supplemented with text descriptions for TV films. Keywords: machine learning, neural networks, themes, topic model, world cloud, LDA-model, linear regression, random forest, decision tree, TVR, predictive models, artificial intelligence, TV advertising, Programmatic platform.