Question

我试图让我的Levenshtein编辑距离算法工作但由于某种原因，编辑的数量不正确。我无法看出我的错误在哪里，我想知道是否有人看到我做错了什么。

输入

id

预期产出

5
ATCGTT
AGTTAC
ACGAAT
CCGTAAAT
TTACGACCAGT

我的输出

Strand A: ATCGTT--
Strand B: A--GTTAC
Edit Distance: 4

Strand A: ATCG-TT
Strand B: A-CGAAT
Edit Distance: 3

Strand A: ATCGT---T
Strand B: -CCGTAAAT
Edit Distance: 5

Strand A: AT-CG----TT
Strand B: TTACGACCAGT
Edit Distance: 7

Strand A: AGTTAC
Strand B: ACGAAT
Edit Distance: 4

Strand A: -AGT-TAC
Strand B: CCGTAAAT
Edit Distance: 5

Strand A: --A-G-TTA-C
Strand B: TTACGACCAGT
Edit Distance: 8

Strand A: ACG--AAT
Strand B: CCGTAAAT
Edit Distance: 3

Strand A: --ACGA--A-T
Strand B: TTACGACCAGT
Edit Distance: 5

Strand A: --CCG-TAAAT
Strand B: TTACGACCAGT
Edit Distance: 7

findEditDistance

Strand A: ATCGT-
Strand B: AGTTAC
Edit Distance: 5

Strand A: ATC-T-
Strand B: ACGAAT
Edit Distance: 5

Strand A: ATC-T-
Strand B: CCGTAAAT
Edit Distance: 5

Strand A: A-C-T-
Strand B: TTACGACCAGT
Edit Distance: 10

Strand A: AGTTAC
Strand B: ACGAAT
Edit Distance: 5

Strand A: AG-TAC
Strand B: CCGTAAAT
Edit Distance: 6

Strand A: A--T-C
Strand B: TTACGACCAGT
Edit Distance: 7

Strand A: AC-AAT
Strand B: CCGTAAAT
Edit Distance: 7

Strand A: AC---T
Strand B: TTACGACCAGT
Edit Distance: 8

Strand A: CC-TAAAT
Strand B: TTACGACCAGT
Edit Distance: 8

getDistance的

void EditDistance::findEditDistance()
{
    int upperValue, leftValue, diagonalValue;

    for (int i = 0; i < mLengthX; ++i)
    {
        table[i][0].stringLength = i;
    }

    for (int i = 0; i < mLengthY; ++i)
    {
        table[0][i].stringLength = i;
    }

    for (int i = 1; i < mLengthX; ++i)
    {
        for (int j = 1; j < mLengthY; ++j)
        {
            if (mStringX[i] == mStringY[j])
            {
                table[i][j].direction = DIAGONAL;
                table[i][j].stringLength = table[i - 1][j -1].stringLength;
            }
            else
            {
                upperValue = table[i - 1][j].stringLength;
                leftValue = table[i][j - 1].stringLength;
                diagonalValue = table[i - 1][j - 1].stringLength;

                if (upperValue < leftValue)
                {
                    if (upperValue < diagonalValue)
                    {
                        //upper is the lowest
                        table[i][j].stringLength = table[i - 1][j].stringLength + 1;
                        table[i][j].direction = UP;
                    }
                    else
                    {
                        //diagonal is lowest
                        table[i][j].stringLength = table[i - 1][j -1].stringLength + 1;
                        table[i][j].direction = DIAGONAL;
                    }
                }
                else if (leftValue < diagonalValue)
                {
                    //left is lowest
                    table[i][j].stringLength = table[i][j - 1].stringLength + 1;
                    table[i][j].direction = LEFT;
                }
                else
                {
                    //diagonal is lowest
                    table[i][j].stringLength = table[i - 1][j -1].stringLength + 1;
                    table[i][j].direction = DIAGONAL;
                }
            }
        }
    }   
}

updateStrands

void EditDistance::getDistance()
{
    int i = mStringX.length() - 1;
    int j = mStringY.length() - 1;

    numEdits = 0;

    updateStrands (i, j);
}

Answer 1

编辑距离的问题出在updateStrands。它将对角线移动计为1，而实际上对角移动的距离可以是1（替换）或0（匹配）。您可以在updateStrands中修复此问题，但当数字已经位于table的右下角时，根本不需要在那里进行计算。

如果你想要正确的＆＃34; strands＆＃34; （例如＆＃34; ATCGTT - ＆＃34;和＃34; A - GTTAC＆＃34;），您必须在updateStrands（您更改元素）中进行更正你应该插入）的字符串，getDistance（你从错误的地方开始）和findEditDistance（你忽略了在上面为direction赋值将stringLength设置为i时的左边缘。

Levenshtein编辑距离不计算编辑距离

1 个答案: