73 lines
3.0 KiB
Python
73 lines
3.0 KiB
Python
|
import pandas as pd
|
|||
|
|
import matplotlib.pyplot as plt
|
|||
|
|
from sklearn.feature_extraction.text import TfidfVectorizer
|
|||
|
|
from sklearn.preprocessing import LabelEncoder
|
|||
|
|
from sklearn.cluster import DBSCAN
|
|||
|
|
from sklearn.decomposition import PCA
|
|||
|
|
from sklearn.metrics import silhouette_score
|
|||
|
|
from sklearn.preprocessing import StandardScaler
|
|||
|
|
from sklearn.preprocessing import normalize
|
|||
|
|
from collections import Counter
|
|||
|
|
|
|||
|
|
filename = "mobiles.csv"
|
|||
|
|
# Считываем данные из файла в DataFrame
|
|||
|
|
data = pd.read_csv(filename, sep=',')
|
|||
|
|
# Удаляем столбец с идентификатором
|
|||
|
|
data.pop("Id")
|
|||
|
|
|
|||
|
|
# Приведение строковых значений признаков к численным при помощи векторайзера с суммированием
|
|||
|
|
FEATURE_COLUMNS_TO_PROC = ['Name', 'Brand', 'Model', 'Operating system']
|
|||
|
|
for column_name in FEATURE_COLUMNS_TO_PROC:
|
|||
|
|
vectorizer = TfidfVectorizer()
|
|||
|
|
train_text_feature_matrix = vectorizer.fit_transform(data[column_name]).toarray()
|
|||
|
|
a = pd.DataFrame(train_text_feature_matrix)
|
|||
|
|
data[column_name] = a[a.columns[1:]].apply(lambda x: sum(x.dropna().astype(float)), axis=1)
|
|||
|
|
|
|||
|
|
# Приведение строковых значений к численным при помощи числового кодирования LabelEncoder
|
|||
|
|
le = LabelEncoder()
|
|||
|
|
data['Touchscreen'] = le.fit_transform(data['Touchscreen'])
|
|||
|
|
data['Wi-Fi'] = le.fit_transform(data['Wi-Fi'])
|
|||
|
|
data['Bluetooth'] = le.fit_transform(data['Bluetooth'])
|
|||
|
|
data['GPS'] = le.fit_transform(data['GPS'])
|
|||
|
|
data['3G'] = le.fit_transform(data['3G'])
|
|||
|
|
data['4G/ LTE'] = le.fit_transform(data['4G/ LTE'])
|
|||
|
|
|
|||
|
|
scaler = StandardScaler()
|
|||
|
|
X_scaled = scaler.fit_transform(data)
|
|||
|
|
# Нормализация масштабированных данных
|
|||
|
|
X_normalized = normalize(X_scaled)
|
|||
|
|
X_normalized = pd.DataFrame(X_normalized)
|
|||
|
|
|
|||
|
|
pca = PCA(n_components=2)
|
|||
|
|
X_principal = pca.fit_transform(X_normalized)
|
|||
|
|
X_principal = pd.DataFrame(X_principal)
|
|||
|
|
X_principal.columns = ['P1', 'P2']
|
|||
|
|
|
|||
|
|
# Определяем и обучаем модель
|
|||
|
|
dbscan = DBSCAN(eps=0.05, min_samples=5).fit(X_principal)
|
|||
|
|
labels = dbscan.labels_
|
|||
|
|
|
|||
|
|
# Вычисление количества кластеров
|
|||
|
|
N_clus = len(set(labels))-(1 if -1 in labels else 0)
|
|||
|
|
print('Количество получившихся кластеров: %d' % N_clus)
|
|||
|
|
counter = Counter(dbscan.labels_)
|
|||
|
|
clusters_df = pd.DataFrame({'Номер кластера': counter.keys(), 'Кол-во элементов': counter.values()}) \
|
|||
|
|
.sort_values(by='Кол-во элементов', ascending=False)
|
|||
|
|
print(clusters_df.reset_index(drop=True))
|
|||
|
|
|
|||
|
|
# Вычисление количества шумовых точек
|
|||
|
|
n_noise = list(dbscan.labels_).count(-1)
|
|||
|
|
print('Количество шумовых точек: %d' % n_noise)
|
|||
|
|
|
|||
|
|
# Оценка качества кластеризации с помощью silhouette_score
|
|||
|
|
silhouette_avg = silhouette_score(X_principal, labels)
|
|||
|
|
|
|||
|
|
print("Silhouette Score:", silhouette_avg)
|
|||
|
|
|
|||
|
|
plt.scatter(X_principal['P1'], X_principal['P2'], c=labels)
|
|||
|
|
plt.savefig('dbscan_plot.png')
|
|||
|
|
# plt.show()
|
|||
|
|
|
|||
|
|
|
|||
|
|
|