1type Ratings = Record<string, Record<string, number>>;
2
3function euclidOnOverlap(a: Record<string, number>, b: Record<string, number>): number {
4  const common = Object.keys(a).filter(k => k in b); 
5  // common – это массив общих ключей двух объектов a и b, то есть пересечение их ключей.
6  if (common.length === 0) return Infinity;
7  let sumSq = 0; // sumSq – это накопитель суммы квадратов разностей оценок на пересечении ключей.
8  for (const k of common) sumSq += (a[k] - b[k]) ** 2;
9  return Math.sqrt(sumSq);
10}
11
12/**
13 * Рекомендательная система k-NN.
14 * 1) Ищем k ближайших пользователей (схожесть = 1 / (1 + distance)).
15 * 2) Для фильмов, которых у user нет, считаем взвешенное среднее рейтингов соседей.
16 */
17function knnRecommend(data: Ratings, user: string, k: number): {
18  neighbors: string[];
19  recommendations: Array<{ item: string; score: number }>;
20} {
21  const dists: Array<{ name: string; dist: number; sim: number }> = [];
22  for (const [name, r] of Object.entries(data)) {
23    if (name === user) continue;
24    const d = euclidOnOverlap(data[user], r); // d – евклидова дистанция
25    if (!Number.isFinite(d)) continue;
26    dists.push({ name, dist: d, sim: 1 / (1 + d) });
27  }
28  dists.sort((a, b) => a.dist - b.dist);
29  const neighbors = dists.slice(0, k);
30
31  const seen = new Set(Object.keys(data[user]));
32  const score: Record<string, number> = {};
33  const weight: Record<string, number> = {};
34
35  for (const n of neighbors) {
36    for (const [item, rating] of Object.entries(data[n.name])) {
37      if (seen.has(item)) continue;
38      score[item]  = (score[item]  ?? 0) + n.sim * rating;
39      weight[item] = (weight[item] ?? 0) + n.sim;
40    }
41  }
42
43  const recommendations = Object.keys(score)
44    .map(item => ({ item, score: +(score[item] / weight[item]).toFixed(2) }))
45    .sort((a, b) => b.score - a.score);
46
47  return { neighbors: neighbors.map(n => n.name), recommendations };
48}
49
50// Данные
51const ratings: Ratings = {
52  you:   { Inception: 5, Matrix: 4 },
53  alice: { Inception: 5, Matrix: 5, Titanic: 1 },
54  bob:   { Inception: 4, Matrix: 4, Avatar: 3, Titanic: 1 },
55  carol: { Matrix: 5, Interstellar: 5, Avatar: 2 },
56  dave:  { Inception: 2, Titanic: 5, Avatar: 4 },
57  eve:   { Matrix: 4, Interstellar: 4, Inception: 5 },
58};
59
60// Примеры
61knnRecommend(ratings, 'you', 2); 
62    // neighbors -> ["eve","alice"]; 
63    // top -> [["Interstellar",4],["Titanic",1]]
64knnRecommend(ratings, 'you', 3); 
65    // neighbors -> ["eve","alice","bob"]; 
66    // top -> [["Interstellar",4],["Avatar",3],["Titanic",1]]
67knnRecommend(ratings, 'you', 4); 
68    // neighbors -> ["eve","alice","bob","carol"]; 
69    // top -> [["Interstellar",4.33],["Avatar",2.5],["Titanic",1]]