saturn

This dataset captures 71 five-minute snapshots of auroral activity on January 20, 2026, with each row recording a timestamp, activity classification, intensity, and power readings for the northern and southern hemispheres. The most striking feature is that 'Storm' conditions dominate 55% of the observations, with 'Active' and 'Quiet' states making up the remainder — suggesting this day saw sustained geomagnetic disturbance. Both north and south power readings show wide, roughly uniform distributions (IQR ~110 GW) with medians well below their means, hinting that storm periods drive the upper range of power values. Intensity is similarly spread across most of its 0–1 range with near-zero skew, making the relationship between activity class and intensity worth exploring closely.

This column contains datetime strings representing regular 5-minute interval timestamps on 2026-01-20, making it a time-series index. All 71 values are unique (cardinality 71, null_rate 0.0, top_rate 0.014) and entropy_ratio is effectively 1.0, confirming every row maps to a distinct timestamp. The 'long_tail' alert is misleading here — the distribution is perfectly uniform (each value appears exactly once), not skewed. This column should be parsed as a proper datetime and used as a time index rather than a categorical feature.

This column represents a normalized intensity measure, bounded between 0.11 and 1.0, consistent with a scaled or clipped continuous score. The distribution is notably platykurtic (kurtosis -1.73), indicating a very flat, spread-out distribution rather than a bell curve — values are nearly uniformly scattered across the range. Despite 71 rows, there are only 35 unique values, suggesting the data originates from a discrete or quantized source (e.g., rounded measurements or a fixed rating scale). Skew is negligible (-0.07), and the IQR of 0.79 spans almost the entire range, confirming broad dispersion with no outliers.

This column appears to measure a directional power reading (northward component) for 71 observations, likely a physical or sensor-derived quantity given its name and continuous numeric nature. The distribution is notably platykurtic (kurtosis -1.28), meaning values are spread very flatly across the range of 11–188 with no heavy tails and no outliers detected. The IQR of 110.5 is nearly as large as the full range, and the median (60.0) sits well below the mean (77.27), suggesting a modest right skew (0.45) driven by a cluster of higher values. With 60 unique values across 71 rows and zero nulls or zeros, the data is dense and well-populated but has some repeated measurements worth investigating.

This column likely represents a power measurement (e.g., watts, kilowatts, or a similar energy metric) associated with a southern-facing sensor, panel, or zone. With a range of 11–190 and a mean of 75.77, the distribution is notably flat: an IQR of 113.0 spanning nearly the full range, combined with a platykurtic kurtosis of -1.29, indicates values are spread broadly and uniformly rather than clustering around a central tendency. The median (51.0) sits well below the mean (75.77), confirming modest right skew (0.49), but no outliers are flagged, suggesting this spread is genuine variability rather than contamination. 61 unique values across 71 rows means some repeated readings exist, which may warrant inspection for duplicate observations.

This column represents a categorical activity-level classification, likely describing geophysical or meteorological states, with exactly three levels: 'Storm', 'Active', and 'Quiet'. The dominant class is 'Storm' at 54.9% (39 of 71 rows), which is mildly surprising given that 'Storm' might intuitively be expected as a rarer, extreme condition. The distribution is moderately imbalanced — 'Quiet' accounts for only 12 observations — which may affect model performance on minority classes. Entropy ratio of 0.898 indicates the distribution is reasonably spread but not uniform.