COMPARATIVE EVALUATION OF NORMALIZATION METHODS FOR NUMERIC AND INTEGER DATA IN NETWORK DATABASE SYSTEMS

Abstract

Numeric and integer data constitute the most frequently encountered attribute types in network database systems, appearing in domains ranging from financial transactions and sensor measurements to user identifiers and event counters. Despite their apparent simplicity, these data types present non-trivial normalization challenges: numeric attributes may span multiple orders of magnitude, contain outliers, or follow non-Gaussian distributions, while integer attributes may be nominal, ordinal, or ratio-scaled, each requiring a different treatment. This paper provides a focused comparative evaluation of normalization methods applicable to numeric and integer data in network databases, examining min-max scaling, Z-score standardization, decimal scaling, robust scaling, and interval-based encoding. For each method, we analyze the mathematical formulation, output range, sensitivity to outliers, distributional assumptions, and suitability for downstream tasks including machine learning, ontological mapping, and network visualization. We further propose a decision framework for selecting the appropriate method based on data characteristics. Experimental validation is conducted on a synthetic network database of 1,000 records, demonstrating that method selection has a measurable impact on data quality metrics and downstream analytical consistency.