102 lines
4.0 KiB
Python
102 lines
4.0 KiB
Python
import pandas as pd
|
||
from sklearn.ensemble import RandomForestRegressor
|
||
from sklearn.metrics import mean_squared_error
|
||
from sklearn.model_selection import train_test_split
|
||
from sklearn.preprocessing import LabelEncoder
|
||
from sqlalchemy import create_engine
|
||
|
||
|
||
def run_prediction_module():
|
||
engine = create_engine('mysql+pymysql://wind:wind@193.124.203.110:3306/wind_towers')
|
||
|
||
query = """
|
||
SELECT BarTrend, CRC, DateStamp, DewPoint, HeatIndex, ETDay, HumIn, HumOut,
|
||
Pressure, RainDay, RainMonth, RainRate, RainStorm, RainYear,
|
||
TempIn, TempOut, WindDir, WindSpeed, WindSpeed10Min
|
||
FROM weather_data
|
||
WHERE DateStamp >= '2024-10-14 21:00:00' - INTERVAL 36 HOUR;
|
||
"""
|
||
df = pd.read_sql(query, engine)
|
||
|
||
df['DateStamp'] = pd.to_datetime(df['DateStamp'])
|
||
df.set_index('DateStamp', inplace=True)
|
||
df.sort_index(inplace=True)
|
||
|
||
lags = 3
|
||
shifted_dfs = [df]
|
||
|
||
for lag in range(1, lags + 1):
|
||
shifted_df = df.shift(lag).add_suffix(f'_t-{lag}')
|
||
shifted_dfs.append(shifted_df)
|
||
|
||
df_with_lags = pd.concat(shifted_dfs, axis=1)
|
||
|
||
df_with_lags.dropna(inplace=True)
|
||
df_with_lags = df_with_lags.copy()
|
||
|
||
# Преобразуем BarTrend в числовой формат
|
||
le = LabelEncoder()
|
||
df_with_lags['BarTrend_encoded'] = le.fit_transform(df_with_lags['BarTrend'])
|
||
|
||
# Выбор только числовых данных
|
||
df_with_lags = df_with_lags.select_dtypes(include=['float64', 'int64'])
|
||
|
||
# Словари для хранения моделей и MSE
|
||
models = {}
|
||
mse_scores = {}
|
||
|
||
# Обучение моделей для каждого целевого столбца
|
||
for target_column in df.columns:
|
||
if target_column not in df_with_lags.columns:
|
||
continue
|
||
|
||
X = df_with_lags.drop(columns=[target_column]).values
|
||
y = df_with_lags[target_column].values
|
||
|
||
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, shuffle=False)
|
||
|
||
model = RandomForestRegressor()
|
||
model.fit(X_train, y_train)
|
||
|
||
y_pred = model.predict(X_test)
|
||
mse = mean_squared_error(y_test, y_pred)
|
||
mse_scores[target_column] = mse
|
||
models[target_column] = model
|
||
|
||
quality = "хорошая" if mse < 1.0 else "плохая"
|
||
print(f"MSE для {target_column}: {mse} ({quality})")
|
||
|
||
# Обучаем модель для BarTrend_encoded отдельно
|
||
X_bartrend = df_with_lags.drop(columns=['BarTrend_encoded']).values
|
||
y_bartrend = df_with_lags['BarTrend_encoded'].values
|
||
|
||
X_train_bartrend, X_test_bartrend, y_train_bartrend, y_test_bartrend = train_test_split(X_bartrend, y_bartrend,
|
||
test_size=0.2,
|
||
shuffle=False)
|
||
|
||
model_bartrend = RandomForestRegressor()
|
||
model_bartrend.fit(X_train_bartrend, y_train_bartrend)
|
||
|
||
y_pred_bartrend = model_bartrend.predict(X_test_bartrend)
|
||
mse_bartrend = mean_squared_error(y_test_bartrend, y_pred_bartrend)
|
||
models['BarTrend_encoded'] = model_bartrend
|
||
mse_scores['BarTrend_encoded'] = mse_bartrend
|
||
|
||
quality_bartrend = "хорошая" if mse_bartrend < 1.0 else "плохая"
|
||
print(f"MSE для BarTrend: {mse_bartrend} ({quality_bartrend})")
|
||
|
||
last_data = X[-1].reshape(1, -1)
|
||
|
||
predictions = {}
|
||
for target_column, model in models.items():
|
||
prediction = model.predict(last_data)[0]
|
||
if target_column == 'BarTrend_encoded':
|
||
prediction = le.inverse_transform([int(prediction)])[0]
|
||
predictions['BarTrend'] = prediction
|
||
print(f"Предсказание для BarTrend: {prediction}")
|
||
break
|
||
predictions[target_column] = prediction
|
||
print(f"Предсказание для {target_column}: {prediction}")
|
||
|
||
return predictions # Возвращаем словарь с предсказанными значениями и названиями столбцов
|