.NET Implementation of the relatively new T-Digest quantile estimation algorithm. Useful for calculating highly accurate Quantiles or Percentiles from on-line streaming data, or data-sets that are too large to store in memory and sort, which is required to calculate the true quantile.
Fully refactored fork of quantumtunneling/T-Digest.NET with next changes:
- Modern .NET frameworks support
- Highly improved performance
- A bit better accuracy
The Nuget package for this Implementation can be found here
The T-Digest white paper can be found here
Example Code:
using TDigestNet;
...
Random random = new();
TDigest digest = new();
for (int i = 0; i < 1000000; i++)
{
var n = random.NextDouble() * 100;
digest.Add(n);
}
Console.WriteLine($"Average: {digest.Average}");
Console.WriteLine($"Percentile 10: {digest.Quantile(10 / 100d)}");
Console.WriteLine($"Percentile 50: {digest.Quantile(50 / 100d)}");
Console.WriteLine($"Percentile 80: {digest.Quantile(80 / 100d)}");
Console.WriteLine($"Percentile 99: {digest.Quantile(99 / 100d)}");Copyright (C) 2023 Aleksej Solomatin