Метод оценки эффективности мероприятий по формированию комфортной городской среды

Abstract

The article deals with the urgent problem of inconsistency of the implemented measures to form a comfortable urban environment with the real needs and expectations of citizens. It analyzes the existing methods of assessing such measures, including the index method of assessing the quality of urban environment, used within the framework of the national project “Housing and Urban Environment”, as well as va¬rious sociological studies and surveys. The shortcomings of these approaches, such as the limited sample of respondents and survey topics, are noted. The aim of the study is to develop a method for evaluating measures to form a comfortable urban environment based on semantic comparison of citizens' opinions from social networks, search queries and descriptions of the measures themselves using natural language processing algorithms. Materials and methods. To realize the set goal, an integrated approach consisting of four main stages is used. Stage 1: Pre-processing of initial text data – noise removal, reduction of words to their initial form (lemmatization) using the pymorphy2 library, identification of parts of speech (POS-tagging). Stage 2: Extraction of key word combinations (N-grams) using the TF-IDF algorithm, taking into account the frequency of usage within individual messages and in the whole text array. Calculation of N-grams significance rank. Stage 3: Obtaining a vector representation (word embeddings) for each key N-gram using the pre-trained SBERT neural network model. Step 4: Computing a measure of semantic similarity of vector representations of N-grams from different text arrays (citizens' opinions and event descriptions) based on cosine distance. The texts of citizens' opinions and descriptions of measures aimed at the formation of a comfortable urban environment related to the Khanty-Mansi Autonomous Okrug are used as input data. Results. The experiment has shown that most of the analyzed measures aimed at the formation of a comfortable urban environment are poorly correlated with the real needs of citizens. Conclusion. The proposed method can be used in decision support systems to evaluate and select the most effective measures.