Уровень 0 · материалов: 3
Сюда входят документы о применении конкретных статистических инструментов и методов обработки числовых данных для анализа или подготовки признаков.
Общие признаки: обработка данных, статистические метрики, нормализация данных, работа с выбросами
Группа выше: Статистический анализ данных
Смысл: The main idea is to demystify percentiles for engineers, explaining how they serve as a tool to condense large datasets into single, meaningful numbers that describe the behavior of the majority of a system while ignoring misleading outliers.
An accessible guide explaining how percentiles work in systems monitoring, specifically focusing on their application in Prometheus via Histograms and Summaries.
Смысл: The main idea is that traditional data normalization methods often fail when dealing with non-normal distributions and outliers, leading to biased feature influence. The author proposes a more robust approach using median centering and a skewness-adjusted interval for scaling to ensure mathematical fairness across all data features.
The author proposes using median centering and a skewness-adjusted interval scaling method to overcome the biases and outlier sensitivity inherent in standard data normalization techniques.
Смысл: The main idea is to explain the mathematical and practical transition of the Gini coefficient from a measure of wealth distribution in economics to a normalized metric for evaluating the predictive power of binary classification models in machine learning.
An educational guide explaining the Gini coefficient's evolution from an economic inequality measure to a machine learning metric, proving its relationship to AUC-ROC and demonstrating its use in scoring and marketing.