time_series_forecasting.py

import pandas as pd
import numpy as np
from statsmodels.tsa.arima.model import ARIMA
from statsmodels.tsa.seasonal import seasonal_decompose
from prophet import Prophet

class TimeSeriesForecaster:
    def __init__(self):
        self.model = None

    def forecast(self, data, date_column, value_column, periods=30, method='auto'):
        # Ensure data is sorted by date
        data = data.sort_values(date_column)
        data = data.set_index(date_column)
        
        if method == 'auto':
            # Automatically choose between ARIMA and Prophet based on data characteristics
            if self._has_seasonality(data[value_column]):
                method = 'prophet'
            else:
                method = 'arima'

        if method == 'arima':
            return self._forecast_arima(data[value_column], periods)
        elif method == 'prophet':
            return self._forecast_prophet(data.reset_index(), date_column, value_column, periods)
        else:
            raise ValueError("Invalid method. Choose 'arima', 'prophet', or 'auto'.")

    def _has_seasonality(self, series, threshold=0.1):
        result = seasonal_decompose(series, model='additive', extrapolate_trend='freq')
        return np.abs(result.seasonal).mean() > threshold * np.abs(result.trend).mean()

    def _forecast_arima(self, series, periods):
        model = ARIMA(series, order=(1, 1, 1))
        self.model = model.fit()
        forecast = self.model.forecast(steps=periods)
        return pd.DataFrame({'date': forecast.index, 'forecast': forecast.values})

    def _forecast_prophet(self, df, date_column, value_column, periods):
        df = df.rename(columns={date_column: 'ds', value_column: 'y'})
        model = Prophet()
        self.model = model.fit(df)
        future = model.make_future_dataframe(periods=periods)
        forecast = model.predict(future)
        return forecast[['ds', 'yhat', 'yhat_lower', 'yhat_upper']]

    def plot_forecast(self, original_data, forecast_data):
        import matplotlib.pyplot as plt

        plt.figure(figsize=(12, 6))
        plt.plot(original_data.index, original_data, label='Original Data')
        plt.plot(forecast_data['date'], forecast_data['forecast'], label='Forecast', color='red')
        plt.fill_between(forecast_data['date'], 
                         forecast_data['forecast'] - forecast_data['forecast'].std(),
                         forecast_data['forecast'] + forecast_data['forecast'].std(),
                         color='red', alpha=0.2)
        plt.legend()
        plt.title('Time Series Forecast')
        plt.xlabel('Date')
        plt.ylabel('Value')
        return plt