ai-analytics/prophet_model.py

from datetime import datetime
import pickle
import pandas as pd
from prophet import Prophet

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

    def train_model(self, csv_file):
        """
        Train the Prophet model using the data from the given CSV file.
        
        :param csv_file: Path to the CSV file containing the dataset.
        """
        # Load the dataset
        df = pd.read_csv(csv_file)

        # Prepare the data
        df.rename(columns={'time': 'ds', 'volume': 'y'}, inplace=True)

        # Initialize and fit the Prophet model
        self.model = Prophet()
        self.model.fit(df)

    def predict(self, datetime):
        """
        Query the forecast for a given date.
        
        :param date: The date for which to query the forecast (format: 'YYYY-MM-DD').
        :return: The forecasted value for the given date.
        """
        if self.model is None:
            raise Exception("Model has not been trained. Call train_model() first.")

        # Create a DataFrame with the given date
        future = pd.DataFrame({'ds': [datetime]})

        # Make predictions
        forecast = self.model.predict(future)

        # Return the forecasted value to json format
        return forecast[['ds', 'yhat', 'yhat_lower', 'yhat_upper']].to_json(orient='records')

    def save_model(self, model_file):
        """
        Save the trained model to a file.
        
        :param model_file: Path to the file where the model will be saved.
        """
        #check if trained model exists
        if self.model is None:
            raise Exception("Model has not been trained. Call train_model() first.")

        with open(model_file, 'wb') as fout:
            pickle.dump(self.model, fout)

    def load_model(self, model_file):
        """
        Load a trained model from a file.
        
        :param model_file: Path to the file containing the saved model.
        """
        with open(model_file, 'rb') as fin:
            self.model = pickle.load(fin)
        return self

# prophet_model = ProphetModel()
# prophet_model.train_model('data/traffic_data.csv')
# print(type(prophet_model))
# predict_time = datetime.strptime('2023-10-01 09:00:00', "%Y-%m-%d %H:%M:%S")
# forecast = prophet_model.predict(predict_time)
# prophet_model.save_model('saved_models/forecast_model.pkl')
# print(forecast)
# prophet_model = prophet_model.load_model('saved-models/forecast_model.pkl')
# print(type(prophet_model))
# forecast = prophet_model.predict(predict_time)
# print(forecast)