PREPROCESSING OF HIERARCHICAL DATA FROM MEDIA PLATFORMS USING RULE-BASED SYSTEMS

Abstract

The article explores preprocessing of hierarchical data from digital advertising platforms using rule-based systems. With rapid growth in digital marketing, companies increasingly rely on structured data analysis from platforms like Meta for decision-making. Advertising data often have complex hierarchical structures organized into campaigns, ad sets, ads, and creatives, each level providing different detail granularity. Efficient preprocessing, including cleaning, transformation, filtering, and integration, is essential to ensure analytical accuracy and optimized budget allocation. Manual preprocessing or general-purpose scripts commonly used are labor-intensive and error-prone, leading to unreliable analytics. To address these challenges, the study proposes a rule-based preprocessing methodology, enabling automated, transparent, and scalable data handling aligned with marketing domain expertise. Rule-based systems use explicit logic ("if-then" rules) to process data systematically, effectively incorporating domain knowledge and reducing errors inherent in manual approaches. A dataset from the Meta platform with nearly 20,000 rows and 232 columns, including many missing values, is examined. Practical preprocessing approaches employed involve aggregating relevant features into logical flags or counts and encoding categorical data through one-hot or label encoding. Missing categorical values were labeled explicitly to distinguish genuine absence clearly from unintentional gaps. These methods targeted data related to geolocation, audience segmentation, creative automation, asset management, tracking pixels, and targeting relaxation. The developed approach significantly reduced dataset dimensionality and eliminated missing values, ensuring data quality suitable for advanced analytics such as clustering or performance evaluation. The rule-based framework proved effective, facilitating standardized, reliable preprocessing critical for robust advertising analysis.