The relevance of the discussed issue is caused by the need to investigate the behavior of test ranking algorithms. The practical value of the research consists in searching for engines developers including the solution of problems of recognition and adaptive classification of objects according to satellite geomonitoring systems. The main aim of the study is to investigate a neural network model of the geomonitoring database text documents ranking algorithm. The model is built on the basis of Kohonen network, multilayer perceptrons, and k-means clustering method. The methods used in the study: software implementation and testing of the neural network ranking algorithms by comparing their work results with the results of the classical ranking algorithm OkapiBm25. The results. The authors have proposed the algorithm, built on the basis of the neural network models of data processing and comprising factor and regression analysis, for the geomonitoring database text retrieval systems identification. Factor analysis includes data clustering based on the use of Kohonen network. To simplify the learning, the factor analysis algorithm is developed to eliminate the characteristics irrelevant to rank. The result the models operation is a set of important ranking characteristics and their optimal values. To perform a regression analysis, it is proposed to use one of two neural network models based on a hybrid neural network or a multilayer perceptrons complex. The regression analysis model is selected on the base of the cluster and factor analysis results. In the case of allocating a large number of the input vectors clusters, a neural network hybrid model is preferable. In the case of the weak intersections between the clusters sets of the significant characteristics, a model based on a set of multilayer perceptrons is preferable. The algorithm testing results show the successful models learning and the low training and testing error values. The proposed models are approved on the OkapiBm25 algorithm's test data, and their application peculiarities are identified depending on the input data characteristics.