60 sec project

# сделать проект за 60 секунд

import pandas as pd
from sklearn.model_selection import train_test_split
from lightautoml.automl.presets.tabular_presets import TabularUtilizedAutoML
from lightautoml.tasks import Task
from sklearn.metrics import mean_squared_error

RANDOM_STATE = 12345 # fixed random state for various reasons

df = pd.read_csv('autos.csv')
X,X_test = train_test_split(df,test_size=0.2,random_state=RANDOM_STATE)

N_FOLDS = 5 # folds cnt for AutoML
TIMEOUT = 60 # Time in seconds for automl run USE TIMEOUT = 1700 for perfect score

task = Task('reg', loss='mse', metric=mean_squared_error, greater_is_better=False)

roles = {
    'target': 'Price',
    # 'drop': ['DateCrawled', 'DateCreated', 'LastSeen', 'NumberOfPictures']
}

automl = TabularUtilizedAutoML(task = task,
                               timeout = TIMEOUT,
                               cpu_limit = 1,
                               random_state = RANDOM_STATE,
                               reader_params = {'n_jobs': 1,
                                                'cv': N_FOLDS,
                                                'random_state': RANDOM_STATE})

oof_pred = automl.fit_predict(X, roles = roles)

preds = automl.predict(X_test).data[:, 0]
result = mean_squared_error(X_test['Price'], preds) ** 0.5
print(f'RMSE: {result}')