Klasterlash

import numpy as np
import pandas as pd
from sklearn.cluster import KMeans
from sklearn.preprocessing import StandardScaler

np.random.seed(42)

mijozlar = pd.DataFrame({
    "yosh":          np.concatenate([
        np.random.normal(25, 4, 100),
        np.random.normal(45, 6, 100),
        np.random.normal(60, 5, 100),
    ]),
    "yillik_xarid":  np.concatenate([
        np.random.normal(2, 0.6, 100),
        np.random.normal(8, 1.5, 100),
        np.random.normal(5, 1.2, 100),
    ]),
    "tashrif_soni":  np.concatenate([
        np.random.normal(30, 8, 100),
        np.random.normal(12, 4, 100),
        np.random.normal(6, 2, 100),
    ]),
})

masshtab = StandardScaler()
X = masshtab.fit_transform(mijozlar)

kmeans = KMeans(n_clusters=3, random_state=42, n_init=10)
mijozlar["guruh"] = kmeans.fit_predict(X)

print(mijozlar.groupby("guruh").mean().round(1))
print(mijozlar["guruh"].value_counts().sort_index())

inersiya = []
k_qiymatlari = range(1, 11)

for k in k_qiymatlari:
    model = KMeans(n_clusters=k, random_state=42, n_init=10)
    model.fit(X)
    inersiya.append(model.inertia_)

for k, qiymat in zip(k_qiymatlari, inersiya):
    print(f"k={k}: {qiymat:8.1f}")

from sklearn.metrics import silhouette_score

for k in range(2, 8):
    model = KMeans(n_clusters=k, random_state=42, n_init=10)
    belgilar = model.fit_predict(X)
    ball = silhouette_score(X, belgilar)
    print(f"k={k}: siluet = {ball:.3f}")

from sklearn.cluster import AgglomerativeClustering
from scipy.cluster.hierarchy import dendrogram, linkage
import matplotlib.pyplot as plt

model = AgglomerativeClustering(n_clusters=3, linkage="ward")
belgilar = model.fit_predict(X)

# Dendrogramma
bogliqlik = linkage(X[:50], method="ward")

plt.figure(figsize=(12, 5))
dendrogram(bogliqlik)
plt.title("Dendrogramma")
plt.xlabel("Namunalar")
plt.ylabel("Masofa")
plt.show()

from sklearn.cluster import DBSCAN

model = DBSCAN(eps=0.5, min_samples=5)
belgilar = model.fit_predict(X)

print(f"Klasterlar: {len(set(belgilar)) - (1 if -1 in belgilar else 0)}")
print(f"Shovqin nuqtalari: {list(belgilar).count(-1)}")

# eps ni tanlash uchun
from sklearn.neighbors import NearestNeighbors

izlovchi = NearestNeighbors(n_neighbors=5).fit(X)
masofalar, _ = izlovchi.kneighbors(X)
beshinchi = np.sort(masofalar[:, 4])

plt.plot(beshinchi)
plt.ylabel("5-qo'shnigacha masofa")
plt.title("eps ni tirsak nuqtasidan tanlang")
plt.show()

from sklearn.cluster import KMeans
import matplotlib.pyplot as plt
import matplotlib.image as mpimg

rasm = mpimg.imread("rasm.jpg")
shakl = rasm.shape

# Har bir piksel - RGB nuqta
piksellar = rasm.reshape(-1, 3)

# 16 ta asosiy rangga qisqartiramiz
kmeans = KMeans(n_clusters=16, random_state=42, n_init=10)
kmeans.fit(piksellar)

yangi_piksellar = kmeans.cluster_centers_[kmeans.labels_]
siqilgan = yangi_piksellar.reshape(shakl).astype("uint8")

print(f"Asl ranglar:     {len(np.unique(piksellar, axis=0))}")
print(f"Siqilgan:        16")

from sklearn.metrics import (
    silhouette_score, calinski_harabasz_score, davies_bouldin_score,
)

belgilar = KMeans(n_clusters=3, random_state=42, n_init=10).fit_predict(X)

print(f"Siluet:           {silhouette_score(X, belgilar):.3f}  (yuqori yaxshi)")
print(f"Calinski-Harabasz: {calinski_harabasz_score(X, belgilar):.1f}  (yuqori yaxshi)")
print(f"Davies-Bouldin:    {davies_bouldin_score(X, belgilar):.3f}  (past yaxshi)")