PANDAS FOR DATA MANAGEMENT IN TARGETING SETUP PROJECTS

Abstract

The article examines the application of the Pandas library as the main tool for managing input, internal, and output data in time series forecasting projects within the field of targeted advertising campaigns. An approach is presented for organizing the process of collecting, processing, and analyzing large volumes of marketing data obtained from CRM systems, advertising platforms, and web analytics. The study outlines the stages of data cleaning from missing values and duplicates, unifying time formats, generating aggregated indicators, and creating new features for subsequent machine learning. Particular attention is paid to the preparation of intermediate metrics such as moving averages, standard deviations, seasonality coefficients, and trend components, which provide a deeper understanding of user behavior over time. For modeling and forecasting audience activity, ARIMA, Prophet, and LSTM algorithms are used, demonstrating varying sensitivity to trends, seasonal fluctuations, and short-term anomalies. Code examples, table fragments, and visualizations of results demonstrate that Pandas provides high flexibility in working with time series, simplifies the creation of data processing pipelines, and enhances the transparency of analytical processes. The results prove the effectiveness of integrating Pandas into marketing decision-making systems, as more accurate forecasting of clicks, impressions, and conversions contributes to budget optimization and increases the profitability of advertising campaigns. A promising direction for further research is the expansion of forecasting models by combining Pandas with deep learning libraries and automating analytical processes. Additionally, it is advisable in future work to develop integrated Pandas-based dashboards for real-time monitoring of forecasting results, which will ensure operational analytics and improve the adaptability of marketing strategies.