# Dataset Evaluation Metrics

Dataset evaluation metrics provide a principled way to judge whether a dataset is fit for a given machine learning or data science task. They encompass (a) descriptive statistics to summarize distributions, central tendency, and dispersion; (b) data quality metrics that assess accuracy, completeness, and consistency; (c) dataset complexity metrics that describe the scale and structure of the data; and (d) class balance metrics that reveal the distribution across target labels. Modern practice also requires explicit attention to bias and fairness, data leakage risk, privacy considerations, and task-aligned evaluation. This record expands traditional categories with enhanced descriptive statistics (including skewness, kurtosis, range, and interquartile range), explicit treatment of missing values and outliers, and practical guidance on reporting thresholds and interpretation. It also clarifies terminology choices (metrics vs measures) and highlights potential conceptual gaps such as bias, representativeness, and leakage that can undermine downstream performance if ignored. The four core categories are described in depth below, followed by policies for reporting, replication, and interpretation, plus a concise glossary of related terms.