ApiRest, Imputacion por moda y KNN_Evelyn Eliana Coaquira Flores

[evelynelian]

import requests import csv import pandas as pd

from sklearn.impute import KNNImputer from sklearn.preprocessing import LabelEncoder

URL = "https://api.mediastack.com/v1/news"

api_key = "2fa83844c4d87b909d954103b2511c1e"

params = { 'sources':'cnn', 'language':'en', 'access_key': api_key } response = requests.get(URL,params=params)

if response.status_code == 200:

news_list = response.json()['data']

with open('news.csv','w',newline='',encoding='utf-8') as csvfile:
    writer = csv.DictWriter(csvfile, fieldnames=['title','description','category','country','url'])
    writer.writeheader()
    for news in news_list:
        writer.writerow({
            'title':news['title'],
            'description':news['description'],
            'category':news['category'],
            'country':news['country'],
            'url':news['url']
        })

    # Guardar en DataFrame
    news_df = pd.DataFrame(news_list)
    news_df = news_df[['title', 'description','category','country', 'url']]  # Filtrar solo columnas necesarias
    print(news_df)  # Para verificar el DataFrame

    #imputacion por moda
    moda = news_df['category'].mode()[0]
    news_df['category'].fillna(moda, inplace=True)

    ###imputacion knn
    # Codificar variables categóricas para que KNN pueda procesarlas
    df_encoded = news_df.apply(LabelEncoder().fit_transform)

    # Aplicar imputación KNN
    imputer = KNNImputer(n_neighbors=5)
    df_imputed = imputer.fit_transform(df_encoded)

    # Convertir de nuevo a DataFrame
    df_imputed = pd.DataFrame(df_imputed, columns=news_df.columns)

    ###Inputacion de datos knn (columnas especificas)
    # Seleccionar solo las columnas que queremos imputar
    columnas_a_imputar = ['category', 'country']

    # Codificar las columnas cualitativas
    label_encoders = {}
    for col in columnas_a_imputar:
        if news_df[col].dtype == 'object':  # Solo columnas categóricas
            le = LabelEncoder()
            news_df[col] = le.fit_transform(news_df[col].astype(str))
            label_encoders[col] = le

    # Inicializar el imputador KNN y aplicar imputación
    imputer = KNNImputer(n_neighbors=3)
    news_df[columnas_a_imputar] = imputer.fit_transform(news_df[columnas_a_imputar])

    # Si quieres revertir la codificación de columnas categóricas
    for col in label_encoders:
        news_df[col] = label_encoders[col].inverse_transform(news_df[col].astype(int))

    print(news_df)

    ###
    ##imputer = KNNImputer(n_neighbors=5)
    ##df_imputed = imputer.fit_transform(news_df)

else: print(f'Error al obtener noticias: ', {response.status_code})