Simulation_blackbox_forecast.ipynb

{"nbformat":4,"nbformat_minor":0,"metadata":{"colab":{"provenance":[{"file_id":"1FTVOecBwJN_8U5_aVgRX7iyaEwJDEBKB","timestamp":1711896174209},{"file_id":"1LoWz8SVuekbL0PQ8kzFdUlW3ZBbopyI1","timestamp":1709914727817},{"file_id":"1iPEyGqoUhv7uloOTbRcjfu9wvVIDU1qI","timestamp":1709908790509}],"toc_visible":true,"mount_file_id":"1FTVOecBwJN_8U5_aVgRX7iyaEwJDEBKB","authorship_tag":"ABX9TyPhi3+s5fnOojeFqVbBLODG"},"kernelspec":{"name":"python3","display_name":"Python 3"},"language_info":{"name":"python"}},"cells":[{"cell_type":"markdown","source":["# Install libraries"],"metadata":{"id":"kIajVPG66xOd"}},{"cell_type":"code","source":["pip install pycaret"],"metadata":{"id":"IrcEEAs0fIUj"},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":["# Import packages"],"metadata":{"id":"h84mInUB66vq"}},{"cell_type":"code","execution_count":2,"metadata":{"id":"0M9PDzjac2kc","executionInfo":{"status":"ok","timestamp":1711908662060,"user_tz":-120,"elapsed":11703,"user":{"displayName":"Kai Horlacher","userId":"17473735662964252342"}}},"outputs":[],"source":["import numpy as np\n","import pandas as pd\n","from pycaret.time_series import *"]},{"cell_type":"markdown","source":["# Read dataset"],"metadata":{"id":"p4vtIiQG7Cdi"}},{"cell_type":"code","source":["file_path = 'mnt/Champagne_Sales.csv'\n","data = pd.read_csv(file_path, parse_dates=['date'], index_col='date')\n","\n","# Renaming the sales data column to 'sales' and defining its datatype as float\n","data.columns = ['sales']\n","data['sales'] = data['sales'].astype(float)\n","\n","# Delete data of 1972, since data are not available for the entire year\n","data = data[data.index.year != 1972]\n","\n","# Show dataframe\n","data.head()"],"metadata":{"id":"yoGzOEp43Vaf"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["# Cut champagne sales data for model training\n","data_cut = data[data.index.year < 1968]\n","data_cut.tail()"],"metadata":{"id":"FeVP_vRBmh7P"},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":["# Simulations of Black Box Forecast"],"metadata":{"id":"RFIBmssff0nK"}},{"cell_type":"code","source":["# Init pycaret setup\n","s = setup(\n","    data=data_cut,\n","    target='sales',\n","    fh=12,\n","    fold = 2,\n","    session_id=42,\n",")"],"metadata":{"id":"-KX4XMp3f9R6"},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":["# Create model"],"metadata":{"id":"79kiPHh4xBLc"}},{"cell_type":"code","source":["# Create ARIMA\n","arima = create_model(\"auto_arima\")"],"metadata":{"id":"Os6-X3Civcvi"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["# Create prophet\n","prophet = create_model(\"prophet\", weekly_seasonality=True, daily_seasonality=True)"],"metadata":{"id":"j57ino2JxGZa"},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":["# Create Predictions"],"metadata":{"id":"N6ZLcEB1xOLp"}},{"cell_type":"code","source":["# Create predictions with ARIMA for 5 years\n","y_pred_arima = predict_model(arima, fh = 60)"],"metadata":{"id":"BfbN29Lkq90n","executionInfo":{"status":"ok","timestamp":1711908886963,"user_tz":-120,"elapsed":581,"user":{"displayName":"Kai Horlacher","userId":"17473735662964252342"}}},"execution_count":10,"outputs":[]},{"cell_type":"code","source":["# Create predictions with prophet for 5 years\n","y_pred_prophet = predict_model(prophet, fh = 60)"],"metadata":{"id":"Q4rEwJrjxmwz","executionInfo":{"status":"ok","timestamp":1711908888195,"user_tz":-120,"elapsed":716,"user":{"displayName":"Kai Horlacher","userId":"17473735662964252342"}}},"execution_count":11,"outputs":[]},{"cell_type":"code","source":["# Map sales data to the black box prediction time series\n","start_date = '1967-01-01'\n","end_date = '1971-12-01'\n","data_filt = data[start_date:end_date]\n","\n","# Create results dataframe\n","result = pd.DataFrame()\n","result.index = data_filt.index\n","result['arima'] = y_pred_arima['y_pred'].values\n","result['prophet'] = y_pred_prophet['y_pred'].values\n","result['sales'] = data_filt['sales'].values\n","result['BlackBox'] = result[['arima', 'prophet']].mean(axis=1)\n","result.drop(columns=['arima', 'prophet'], inplace=True)"],"metadata":{"id":"6MYO--FVx86X","executionInfo":{"status":"ok","timestamp":1711908988781,"user_tz":-120,"elapsed":278,"user":{"displayName":"Kai Horlacher","userId":"17473735662964252342"}}},"execution_count":17,"outputs":[]},{"cell_type":"code","source":["result.head()"],"metadata":{"id":"UFY0kbqwuRMU"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["result.plot()"],"metadata":{"id":"xj_wj5Xos-Z3"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["# Angenommen, Ihr DataFrame heißt df\n","path = 'mnt/BlackBox_forecast_simulation.csv'  # Pfad anpassen, wo die CSV gespeichert werden soll\n","result.to_csv(path)  # index=False, falls Sie den Index nicht mit speichern möchten\n"],"metadata":{"id":"lyj00rv0HjvY"},"execution_count":null,"outputs":[]}]}