Skip to content
TheChatArchiveThe Chat Archive
gpt-3.5-turbo-0301· openaiHugging Face Suzhen/CodeChat

c# 实现一个EXCEL中TREND的最小二乘法函数,可输入bool决定是否支持截距为零

  1. User

    c# 实现一个EXCEL中TREND的最小二乘法函数,可输入bool决定是否支持截距为零

  2. Assistant

    以下是一个简单的C#实现,使用了MathNet库来进行矩阵计算:

    using System;
    using System.Linq;
    using MathNet.Numerics.LinearAlgebra;
    using MathNet.Numerics.LinearAlgebra.Double;
    
    public class LinearRegression{
        public static Vector<double> Trend(Vector<double> x, Vector<double> y, bool includeIntercept = true){
            if(x.Count != y.Count){
                throw new ArgumentException("x and y vectors must have the same length");
            }
            var n = x.Count;
    
            //构建X矩阵
            var X = Matrix<double>.Build.Dense(n, includeIntercept ? 2 : 1);
            for(int i=0; i<n; i++){
                X[i, 0] = includeIntercept ? 1 : 0;
                X[i, includeIntercept ? 1 : 0] = x[i];
            }
    
            //计算最小二乘解
            var XTX_inv = X.Transpose() * X;
            if(includeIntercept){
                XTX_inv[0, 0] = 0; //截距为零时,设矩阵的左上角为0
            }
            var XTY = X.Transpose() * y;
            var beta = XTX_inv.Cholesky().Solve(XTY);
    
            return beta;
        }
    }
    
    //使用示例
    var x = Vector<double>.Build.Dense(new double[] { 1, 2, 3, 4, 5 });
    var y = Vector<double>.Build.Dense(new double[] { 1.2, 1.8, 3.2, 3.9, 5.1 });
    
    //计算带截距的趋势线
    var beta1 = LinearRegression.Trend(x, y);
    Console.WriteLine("带截距的最小二乘解: y = {0} + {1} * x", beta1[0], beta1[1]);
    
    //计算截距为零的趋势线
    var beta2 = LinearRegression.Trend(x, y, false);
    Console.WriteLine("截距为零的最小二乘解: y = {0} * x", beta2[0]);
    

    在这个包含截距的情况下,矩阵X看起来如下:

    1 x1
    1 x2
    1 x3
    1 x4
    1 x5
    

    在截距为零的情况下,X矩阵变为:

    0 x1
    0 x2
    0 x3
    0 x4
    0 x5
    

    在这种情况下,矩阵的左上角设置为0,因为没有拦截。