Байесовский подход к оцениванию факторов, влияющих на положение сайта в результатах поискового запроса

This paper presents building regression model with supervised selectivity to be applied for problem of order estimation of sites in result of web search query. Its basic characteristics are small number of observations but big number of descriptive features. Bayesian approach, where maximum likelihood criteria is set up based on the hierarchic model of general set of objects, is proposed for estimating the regression coefficients of the model. The suggested criteria can eliminate redundant factors and keep meaningful ones to determine the positions of sites within the framework of given web search query. The specific model has been tested and confirmed by simulation and real data.