要使用余弦相似度或Levenshtein距離來(lái)計(jì)算字符串之間的相似度，我們需要對(duì)字符串進(jìn)行一些預(yù)處理。余弦相似度通常用于比較文檔或句子，而Levenshtein距離用于計(jì)算兩個(gè)字符串之間的編輯距離。下面我將分別展示如何使用這兩種方法來(lái)對(duì)字符串?dāng)?shù)組進(jìn)行分組。

使用Levenshtein距離

Levenshtein距離是指將一個(gè)字符串轉(zhuǎn)換為另一個(gè)字符串所需的最少編輯操作次數(shù)（包括插入、刪除和替換字符）。

function levenshteinDistance(a, b) {
    if (a.length === 0) return b.length;
    if (b.length === 0) return a.length;
    const matrix = [];
    // 初始化矩陣
    for (let i = 0; i <= b.length; i++) {
        matrix[i] = [i];
    }
    for (let j = 0; j <= a.length; j++) {
        matrix[0][j] = j;
    }
    // 計(jì)算Levenshtein距離
    for (let i = 1; i <= b.length; i++) {
        for (let j = 1; j <= a.length; j++) {
            if (b.charAt(i - 1) === a.charAt(j - 1)) {
                matrix[i][j] = matrix[i - 1][j - 1];
            } else {
                matrix[i][j] = Math.min(
                    matrix[i - 1][j - 1] + 1, // 替換
                    Math.min(
                        matrix[i][j - 1] + 1, // 插入
                        matrix[i - 1][j] + 1 // 刪除
                    )
                );
            }
        }
    }
    return matrix[b.length][a.length];
}
function groupByLevenshtein(strings, distanceThreshold) {
    const groups = [];
    strings.forEach((str, index) => {
        let grouped = false;
        for (let i = 0; i < groups.length; i++) {
            const group = groups[i];
            if (levenshteinDistance(str, group[0]) <= distanceThreshold) {
                group.push(str);
                grouped = true;
                break;
            }
        }
        if (!grouped) {
            groups.push([str]);
        }
    });
    return groups;
}
// 測(cè)試數(shù)據(jù)
const strings = ["apple", "apply", "appale", "orange", "orango", "banana", "bananb"];
const distanceThreshold = 2;
// 調(diào)用函數(shù)并打印結(jié)果
const groups = groupByLevenshtein(strings, distanceThreshold);
console.log(groups);

使用余弦相似度

余弦相似度通常用于比較文檔或句子，它基于向量空間模型。為了使用余弦相似度，我們需要將字符串轉(zhuǎn)換為向量。這通常涉及到將字符串分割成單詞或字符，并創(chuàng)建一個(gè)表示每個(gè)單詞或字符出現(xiàn)次數(shù)的向量。

function termFrequency(str) {
    const terms = str.split('');
    const frequency = {};
    terms.forEach(term => {
        frequency[term] = (frequency[term] || 0) + 1;
    });
    return frequency;
}
function cosineSimilarity(vecA, vecB) {
    let dotProduct = 0;
    let magnitudeA = 0;
    let magnitudeB = 0;
    for (let key in vecA) {
        dotProduct += vecA[key] * (vecB[key] || 0);
        magnitudeA += vecA[key] * vecA[key];
    }
    for (let key in vecB) {
        magnitudeB += vecB[key] * vecB[key];
    }
    return dotProduct / (Math.sqrt(magnitudeA) * Math.sqrt(magnitudeB));
}
function groupByCosineSimilarity(strings, similarityThreshold) {
    const groups = [];
    const vectors = strings.map(termFrequency);
    vectors.forEach((vector, index) => {
        let grouped = false;
        for (let i = 0; i < groups.length; i++) {
            const group = groups[i];
            if (cosineSimilarity(vector, vectors[group[0]]) >= similarityThreshold) {
                group.push(index);
                grouped = true;
                break;
            }
        }
        if (!grouped) {
            groups.push([index]);
        }
    });
    return groups.map(group => group.map(index => strings[index]));
}
// 測(cè)試數(shù)據(jù)
const strings = ["apple", "apply", "appale", "orange", "orango", "banana", "bananb"];
const similarityThreshold = 0.8; // 余弦相似度的范圍是-1到1，接近1表示更相似
// 調(diào)用函數(shù)并打印結(jié)果
const groups = groupByCosineSimilarity(strings, similarityThreshold)
console.log(groups);