Comparative analysis of machine learning results and regression model of online service user behavior trajectories

The dynamics of behavior of online customers is of great interest to marketers in order to maximize the profit of the store and predict the development of online sales. The most widespread methods of processing statistical data are regression statistical analysis methods, and machine learning methods acquire relevance. The purpose of the study was to predict the behavior of online store users, based on the original data obtained by BigData technologies. The correlation of the factor and the result was evaluated, the presence of a direct linear relationship was shown. Classical regression analysis methods would determine linear regression coefficients, assess their significance, model adequacy, mean absolute and relative approximation errors. The model was trained by machine learning methods, coefficients were determined. Comparative results are presented in the form of a graph. The prediction confidence interval was determined for the significance level α = 0,05. Relevant findings are presented.

1. Digital 2021: главная статистика по России и всему миру [Электронный ресурс]. – URL: https://spark.ru/user/115680/blog/74085/digital2021-glavnaya-statistika-po-rossii-i-vsemu-miru/ (дата обращения: 15.03.2023).

2. Прохорова М.В. Организация работы интернет-магазина: Пособие / М.В. Прохорова, А.Л. Коданина. – 3-е изд. М.: Дашков и К, 2020. – 332 с.

3. Gull M. Customer Behavior Analysis Towards Online Shopping using Data Mining, / M. Gull, A. Pervaiz // 2018 5th International Multi-Topic ICT Conference (IMTIC). – 2018. – pp. 1-5.

4. Шипилова Е.А. Анализ и моделирование траекторий поведения пользователей онлайн-сервисов с использованием платформы RETENTIONEERING / Е.А. Шипилова, Е.Е. Некрылов, Т.В Курчекова // Моделирование систем и процессов: Научно-технический журнал. – 2022. – Т. 15. – Вып. 4. – С. 82-93.